From 8063da50b23b7b9896ae29099d9db171ddf5ca45 Mon Sep 17 00:00:00 2001
From: Thomas Debray <117118104+tdebray123@users.noreply.github.com>
Date: Tue, 8 Oct 2024 16:05:36 +0200
Subject: [PATCH] clean up code
---
_freeze/chapter_11/execute-results/html.json | 4 +-
_freeze/chapter_11/figure-html/fig-hmr4-1.png | Bin 33006 -> 26402 bytes
_freeze/chapter_11/figure-html/fig-hmr5-1.png | Bin 0 -> 67755 bytes
chapter_11.qmd | 90 +++++++++++----
chapter_11_files/figure-html/fig-hmr4-1.png | Bin 33006 -> 26402 bytes
chapter_11_files/figure-html/fig-hmr5-1.png | Bin 0 -> 67755 bytes
docs/chapter_11.html | 103 ++++++++++--------
.../figure-html/fig-hmr4-1.png | Bin 33006 -> 26402 bytes
.../figure-html/fig-hmr5-1.png | Bin 0 -> 67755 bytes
docs/search.json | 4 +-
10 files changed, 134 insertions(+), 67 deletions(-)
create mode 100644 _freeze/chapter_11/figure-html/fig-hmr5-1.png
create mode 100644 chapter_11_files/figure-html/fig-hmr5-1.png
create mode 100644 docs/chapter_11_files/figure-html/fig-hmr5-1.png
diff --git a/_freeze/chapter_11/execute-results/html.json b/_freeze/chapter_11/execute-results/html.json
index 04fb36b..30102fd 100644
--- a/_freeze/chapter_11/execute-results/html.json
+++ b/_freeze/chapter_11/execute-results/html.json
@@ -1,8 +1,8 @@
{
- "hash": "2bb698836ad87a92f4ef8dfe3aada86f",
+ "hash": "5620e0a1aa55df1d350f2fd2eb875fcc",
"result": {
"engine": "knitr",
- "markdown": "---\ntitle: \"Individual Participant Data Meta-analysis of clinical trials and real-world data\"\nauthors: \n - name: Pablo Verde\n affiliations:\n - ref: umcdusseldorf\n - name: Thomas Debray\n orcid: 0000-0002-1790-2719\n affiliations:\n - ref: smartdas\naffiliations:\n - id: smartdas\n name: Smart Data Analysis and Statistics B.V.\n city: Utrecht\n - id: umcdusseldorf\n name: Universitätsklinikum Düsseldorf\n city: Düsseldorf\nformat:\n html:\n toc: true\n number-sections: true\nexecute:\n cache: true\nbibliography: 'https://api.citedrive.com/bib/0d25b38b-db8f-43c4-b934-f4e2f3bd655a/references.bib?x=eyJpZCI6ICIwZDI1YjM4Yi1kYjhmLTQzYzQtYjkzNC1mNGUyZjNiZDY1NWEiLCAidXNlciI6ICIyNTA2IiwgInNpZ25hdHVyZSI6ICI0MGFkYjZhMzYyYWE5Y2U0MjQ2NWE2ZTQzNjlhMWY3NTk5MzhhNzUxZDNjYWIxNDlmYjM4NDgwOTYzMzY5YzFlIn0=/bibliography.bib'\n---\n\n\n\n\n## Introduction\n\n\n## Hierarchical Meta-Regression\nWe illustrate the implementation of hierarchical meta-regression using an example that involves the following data sources:\n\n* Aggregate data from 35 randomized trials investigating the efficacy of adjunctive treatments in managing diabetic foot problems compared with routine care\n* Individual participant data from a prospective cohort study investigating patient and limb survival in patients with diabetic foot ulcers \n\n### Aggregate data\nWe first retrieve the randomized evidence and summarize the treatment effect estimates using a random effects meta-analysis:\n\n\n::: {.cell}\n\n```{.r .cell-code}\nlibrary(dplyr)\nlibrary(jarbes)\nlibrary(metafor)\n\ndata(\"healing\")\n\naddat <- escalc(measure=\"OR\", ai=y_t, bi=n_t-y_t, ci=y_c, di=n_c-y_c, data=healing)\n\nresults.ADJ <- metagen(TE = yi, seTE = sqrt(vi),\n studlab = Study, data = addat,\n sm = \"OR\", \n prediction = TRUE)\n```\n:::\n\n\nThe corresponding forest plot is depicted below. The endpoint is healing without amputations within a period less than or equal to 1 year.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![](chapter_11_files/figure-html/unnamed-chunk-3-1.png){width=960}\n:::\n:::\n\n\nThe random effects meta-analysis yielded a pooled odds ratio of 1.89. However, substantial between-study heterogeneity was found, with $\\tau$ = 0.45.\n\n### Individual participant data\nSubsequently, we retrieve the individual participant data:\n\n\n::: {.cell}\n\n```{.r .cell-code}\ndata(\"healingipd\")\nIPD <- healingipd %>% dplyr::select(healing.without.amp, PAD, neuropathy,\n first.ever.lesion, no.continuous.care, \n male, diab.typ2, insulin, HOCHD, \n HOS, CRF, dialysis, DNOAP, smoking.ever, \n diabdur, wagner.class)\n```\n:::\n\n\nBriefly, these IPD were obtained from a prospective cohort study enrolling consecutive patients with diabetic foot ulcers (DFUs) and without previous major amputation in a single diabetes center between June 1998 and December 1999 [@morbach_long-term_2012]. The baseline characteristics of the study population are summarized below:\n\n\n::: {.cell}\n::: {.cell-output-display}\n\n```{=html}\n
\n\n\n | \nHealing without amputation (N=165) | \nNo healing without amputation (N=95) | \nOverall (N=260) | \n
\n\n\n\nAge (years) | \n | \n | \n | \n
\n\nMean (SD) | \n69.1 (10.9) | \n68.5 (11.0) | \n68.9 (10.9) | \n
\n\nMedian [Min, Max] | \n70.0 [25.0, 90.0] | \n69.0 [36.0, 91.0] | \n70.0 [25.0, 91.0] | \n
\n\nDiabetes duration (years) | \n | \n | \n | \n
\n\nMean (SD) | \n15.9 (10.3) | \n15.9 (11.2) | \n15.9 (10.6) | \n
\n\nMedian [Min, Max] | \n14.0 [1.00, 53.0] | \n14.0 [0, 50.0] | \n14.0 [0, 53.0] | \n
\n\nSex | \n | \n | \n | \n
\n\nFemale | \n71 (43.0%) | \n35 (36.8%) | \n106 (40.8%) | \n
\n\nMale | \n94 (57.0%) | \n60 (63.2%) | \n154 (59.2%) | \n
\n\nEver smoker | \n | \n | \n | \n
\n\nYes | \n97 (58.8%) | \n57 (60.0%) | \n154 (59.2%) | \n
\n\nNo | \n68 (41.2%) | \n38 (40.0%) | \n106 (40.8%) | \n
\n\nDiabetes type 2 | \n | \n | \n | \n
\n\nYes | \n150 (90.9%) | \n79 (83.2%) | \n229 (88.1%) | \n
\n\nNo | \n15 (9.1%) | \n16 (16.8%) | \n31 (11.9%) | \n
\n\nPeripheral arterial disease | \n | \n | \n | \n
\n\nYes | \n82 (49.7%) | \n66 (69.5%) | \n148 (56.9%) | \n
\n\nNo | \n83 (50.3%) | \n29 (30.5%) | \n112 (43.1%) | \n
\n\nNeuropathy | \n | \n | \n | \n
\n\nYes | \n144 (87.3%) | \n80 (84.2%) | \n224 (86.2%) | \n
\n\nNo | \n21 (12.7%) | \n15 (15.8%) | \n36 (13.8%) | \n
\n\nFirst ever lesion | \n | \n | \n | \n
\n\nYes | \n70 (42.4%) | \n44 (46.3%) | \n114 (43.8%) | \n
\n\nNo | \n95 (57.6%) | \n51 (53.7%) | \n146 (56.2%) | \n
\n\nNo continuous care | \n | \n | \n | \n
\n\nYes | \n115 (69.7%) | \n62 (65.3%) | \n177 (68.1%) | \n
\n\nNo | \n50 (30.3%) | \n33 (34.7%) | \n83 (31.9%) | \n
\n\nInsulin dependent | \n | \n | \n | \n
\n\nYes | \n109 (66.1%) | \n65 (68.4%) | \n174 (66.9%) | \n
\n\nNo | \n56 (33.9%) | \n30 (31.6%) | \n86 (33.1%) | \n
\n\nHistory of coronary events (CHD) | \n | \n | \n | \n
\n\nYes | \n31 (18.8%) | \n21 (22.1%) | \n52 (20.0%) | \n
\n\nNo | \n134 (81.2%) | \n74 (77.9%) | \n208 (80.0%) | \n
\n\nHistory of stroke | \n | \n | \n | \n
\n\nYes | \n36 (21.8%) | \n19 (20.0%) | \n55 (21.2%) | \n
\n\nNo | \n129 (78.2%) | \n76 (80.0%) | \n205 (78.8%) | \n
\n\nCharcot foot syndrome | \n | \n | \n | \n
\n\nYes | \n28 (17.0%) | \n24 (25.3%) | \n52 (20.0%) | \n
\n\nNo | \n137 (83.0%) | \n71 (74.7%) | \n208 (80.0%) | \n
\n\nDialysis | \n | \n | \n | \n
\n\nYes | \n3 (1.8%) | \n6 (6.3%) | \n9 (3.5%) | \n
\n\nNo | \n162 (98.2%) | \n89 (93.7%) | \n251 (96.5%) | \n
\n\nDiabetic Neuropathic Osteoarthropathy (DNOAP) | \n | \n | \n | \n
\n\nYes | \n19 (11.5%) | \n10 (10.5%) | \n29 (11.2%) | \n
\n\nNo | \n146 (88.5%) | \n85 (89.5%) | \n231 (88.8%) | \n
\n\nWagner score | \n | \n | \n | \n
\n\n1-2 | \n115 (69.7%) | \n27 (28.4%) | \n142 (54.6%) | \n
\n\n3-4-5 | \n50 (30.3%) | \n68 (71.6%) | \n118 (45.4%) | \n
\n\n
\n
\n```\n\n:::\n:::\n\n\nAs depicted above, IPD are available from 260 patients. Some of these patients have similar characteristics to those enrolled in the randomized trials. However, other patients have comorbidities, where one or more risk factors prevent them to participate in the RCTs due to ethical reasons. For example,\n118 patients have severe ulcer lesions (Wagner score 3 to 5), and 77 patients suffer from severe ulcer lesions and peripheral arterial disease (PAD). The question is: Can we generalize the benefit of adjuvant therapies observed in the RCTs to the subgroups of patients encountered in clinical practice?\n\n### Hierarchical metaregression\nWe first investigate the event rate of patients receiving routine care:\n\n\n::: {.cell}\n::: {.cell-output-display}\n![](chapter_11_files/figure-html/unnamed-chunk-6-1.png){width=960}\n:::\n:::\n\n\nThe forest plot above indicates that the baseline risk in the observational study from Morbach et al. is much higher than most trials.\n\n\nWe fitted an HMR model to the available RWD and published AD: \n\n\n::: {.cell}\n\n```{.r .cell-code}\nset.seed(2022)\n\nAD <- healing %>% dplyr::select(yc = y_c, nc = n_c, \n yt = y_t, nt = n_t, Study = Study)\n\nmx2 <- hmr(data = AD, # Published aggregate data\n two.by.two = FALSE, # \n dataIPD = IPD, # Data frame of the IPD \n re = \"sm\", # Random effects model: \"sm\" scale mixtures \n link = \"logit\", # Link function of the random effects\n sd.mu.1 = 1, # Scale parameter for the prior of mu.1\n sd.mu.2 = 1, # Scale parameter for the prior of mu.2 \n sd.mu.phi = 1, # Scale parameter for the prior of mu.phi \n sigma.1.upper = 5, # Upper bound of the prior of sigma.1 \n sigma.2.upper = 5, # Upper bound of the prior of sigma.2\n sigma.beta.upper = 5, # Upper bound of the prior of sigma.beta\n sd.Fisher.rho = 1.25, # Scale parameter for the prior of rho\n df.estimate = TRUE, # If TRUE the degrees of freedom are estimated\n df.lower = 3, # Lower bound of the df's prior\n df.upper = 10, # Upper bound of the df's prior\n nr.chains = 2, # Number of MCMC chains\n nr.iterations = 10000, # Total number of iterations\n nr.adapt = 1000, # Number of iteration for burnin \n nr.thin = 1) # Thinning rate\n```\n:::\n\n\nWe start our analysis by visualizing the conflict of evidence between the different types of data and study types. The figure below depicts the posterior distribution of $\\mu_{\\phi}$, which is the mean bias of the IPD-NRS compared to the AD-RCTs control groups. With only one IPD-NRS, this parameter is partially\nidentifiable from the data. However, we can expect to learn about this bias parameter \nin a full Bayesian model.\n\n\n\n\n\n::: {.cell}\n\n```{.r .cell-code}\n# Pablo's calculations ...\nmu.phi <- mx2$BUGSoutput$sims.list$mu.phi\nmean(mu.phi)\n```\n\n::: {.cell-output .cell-output-stdout}\n\n```\n[1] 0.7868735\n```\n\n\n:::\n\n```{.r .cell-code}\nsd(mu.phi)\n```\n\n::: {.cell-output .cell-output-stdout}\n\n```\n[1] 1.091779\n```\n\n\n:::\n\n```{.r .cell-code}\n# Bias parameters: \n# mean sd 2.5% 25% 50% 75% 97.5% Rhat n.eff\n# mu.phi 0.79 1.09 -1.35 0.09 0.78 1.48 2.96 1 5000\n\n# Pr(mu.phi>0|Data) ...\n\nPr.mu.phi = sum(mu.phi > 0)/length(mu.phi)\nPr.mu.phi\n```\n\n::: {.cell-output .cell-output-stdout}\n\n```\n[1] 0.7733889\n```\n\n\n:::\n:::\n\n\nThe posterior distribution of $\\mu_{\\phi}$ has a mean of 0.79 and a 95% posterior interval of [-1.35, 2.96]. The posterior probability that $\\mu_{\\phi}$ is greater than zero is 77%, which indicates \nthat in average the IPD-NRS of this cohort present a better prognoses that the AD-RCTs control groups.\nThat means that taking the IPD-NRS results at face value would be misleading if we aim to combine them with a meta-analysis of AD-RCTs.\n\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Posterior sensitivity analysis of bias mean between the RCTs and the IPD-NRS.](chapter_11_files/figure-html/fig-hmr1-1.png){#fig-hmr1 width=1056}\n:::\n:::\n\n\n\n@fig-hmr2 presents the posterior distribution of the weights $w_{i}$ for each study included in the HMR. These posteriors are summarized using a forest plot, where posterior intervals substantially greater than one indicate outliers. One important aspect of the HMR is that those outliers are automatically down-weighted in the analysis.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Posterior distribution of the study weights, illustrated by the median and 95% credible intervals. Studies with posterior weights greater than 1.5, marked in red, are flagged as potential outliers.](chapter_11_files/figure-html/fig-hmr2-1.png){#fig-hmr2 width=576}\n:::\n:::\n\n\n@fig-hmr3 displays the results of the submodel corresponding to the\nIPD-NRS that received only medical routine care. The posteriors of the\nregression coefficients $\\beta_k$ ($k=1,\\dots, 15$) are summarized in a forest plot. This submodel\ndetects risk factors that can reduce the chance of getting healed. We see that\nthe group of patients with a Wagner score greater than 2 have substantially less\nchance of getting healed compared to the group with lower scores. This can also\nbe observed in the group of patients with PAD.\n\nInterestingly, these subgroups of patients that have lower chances of getting\nhealed are underrepresented in the RCTs populations. Therefore, by combining\nIPD-NRS with AD-RCT we can learn new insights about these patients that cannot\nbe learned neither from AD nor from IPD alone.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Posterior distribution of regression coefficients from the IPD-NRS analysis, illustrated by the mean and 95% credible intervals. The most relevant risk factors identified in this analysis were the Wagner classification (1-2 vs. 3-4-5) and the presence of peripheral arterial disease (PAD) (no vs. yes).](chapter_11_files/figure-html/fig-hmr3-1.png){#fig-hmr3 width=768}\n:::\n:::\n\n\nThe association between baseline healing risk without amputation within one year and the relative treatment effect is illustrated in @fig-hmr4. Results from the underlying HMR submodel are used to predict treatment effects across different patient subgroups, providing insights into how baseline risk impacts the effectiveness of the treatment. The posterior median and 95\\% credible intervals indicate that healthier patients (with a) are associated with a reduced treatment effect. In other words, healthier patients tend to derive less benefit from the adjunctive therapy compared to those with a higher baseline risk.\n\nThe model is centered at -0.565, corresponding to the posterior mean of $\\mu_1$, the RCTs' baseline risk. To the right of $\\mu_1$ we have the posterior mean of the IPD-NRS $\\mu_1 +\\mu_{\\phi}$, which has a posterior mean of 0.222. This shows an important bias captured by the introduction of $\\mu_{\\phi}$ in the model.\n\n\n::: {.cell}\n::: {.cell-output .cell-output-stderr}\n\n```\nWarning in prop.test(x = healingplus$y_c[i], n = healingplus$n_c[i], correct =\nFALSE): Chi-squared approximation may be incorrect\n```\n\n\n:::\n\n::: {.cell-output .cell-output-stderr}\n\n```\nWarning: Duplicated aesthetics after name standardisation: colour\n```\n\n\n:::\n\n::: {.cell-output .cell-output-stderr}\n\n```\nWarning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.\nℹ Please use `linewidth` instead.\n```\n\n\n:::\n\n::: {.cell-output-display}\n![Summary results of generalizing relative treatment effects: The RCTs' results are displayed as a forest plot. The fitted hierarchical meta-regression model is summarized with solid lines representing the posterior median and 95% intervals. The predicted log odds ratio for the observational study is displayed in blue.](chapter_11_files/figure-html/fig-hmr4-1.png){#fig-hmr4 width=768}\n:::\n:::\n\n\n## Version info {.unnumbered}\nThis chapter was rendered using the following version of R and its packages:\n\n\n::: {.cell}\n::: {.cell-output .cell-output-stdout}\n\n```\nR version 4.4.1 (2024-06-14 ucrt)\nPlatform: x86_64-w64-mingw32/x64\nRunning under: Windows 10 x64 (build 19045)\n\nMatrix products: default\n\n\nlocale:\n[1] LC_COLLATE=English_United Kingdom.utf8 \n[2] LC_CTYPE=English_United Kingdom.utf8 \n[3] LC_MONETARY=English_United Kingdom.utf8\n[4] LC_NUMERIC=C \n[5] LC_TIME=English_United Kingdom.utf8 \n\ntime zone: Europe/Paris\ntzcode source: internal\n\nattached base packages:\n[1] stats graphics grDevices utils datasets methods base \n\nother attached packages:\n [1] meta_7.0-0 metadat_1.2-0 table1_1.4.3 tableone_0.13.2 \n [5] dplyr_1.1.4 jarbes_2.2.1 GGally_2.2.1 R2jags_0.8-5 \n [9] rjags_4-15 mcmcplots_0.4.3 coda_0.19-4.1 gridExtra_2.3 \n[13] ggplot2_3.5.1 kableExtra_1.4.0\n\nloaded via a namespace (and not attached):\n [1] tidyselect_1.2.1 viridisLite_0.4.2 farver_2.1.2 \n [4] fastmap_1.2.0 denstrip_1.5.4 CompQuadForm_1.4.3 \n [7] mathjaxr_1.6-0 promises_1.3.0 digest_0.6.36 \n[10] mime_0.12 lifecycle_1.0.4 survival_3.7-0 \n[13] magrittr_2.0.3 compiler_4.4.1 rlang_1.1.4 \n[16] tools_4.4.1 utf8_1.2.4 yaml_2.3.8 \n[19] knitr_1.48 labeling_0.4.3 htmlwidgets_1.6.4 \n[22] plyr_1.8.9 xml2_1.3.6 RColorBrewer_1.1-3 \n[25] abind_1.4-8 miniUI_0.1.1.1 withr_3.0.1 \n[28] purrr_1.0.2 numDeriv_2016.8-1.1 grid_4.4.1 \n[31] fansi_1.0.6 xtable_1.8-4 colorspace_2.1-0 \n[34] scales_1.3.0 MASS_7.3-61 cli_3.6.3 \n[37] survey_4.4-2 rmarkdown_2.28 metafor_4.6-0 \n[40] generics_0.1.3 rstudioapi_0.16.0 tzdb_0.4.0 \n[43] minqa_1.2.7 DBI_1.2.3 stringr_1.5.1 \n[46] splines_4.4.1 parallel_4.4.1 mitools_2.4 \n[49] vctrs_0.6.5 boot_1.3-31 Matrix_1.7-0 \n[52] jsonlite_1.8.8 hms_1.1.3 Formula_1.2-5 \n[55] systemfonts_1.1.0 tidyr_1.3.1 R2WinBUGS_2.1-22.1 \n[58] glue_1.7.0 nloptr_2.1.1 codetools_0.2-20 \n[61] ggstats_0.6.0 stringi_1.8.4 gtable_0.3.5 \n[64] later_1.3.2 sfsmisc_1.1-19 lme4_1.1-35.5 \n[67] munsell_0.5.1 tibble_3.2.1 pillar_1.9.0 \n[70] htmltools_0.5.8.1 ggExtra_0.10.1 R6_2.5.1 \n[73] evaluate_0.24.0 shiny_1.9.1 lattice_0.22-6 \n[76] readr_2.1.5 httpuv_1.6.15 Rcpp_1.0.13 \n[79] svglite_2.1.3 nlme_3.1-165 xfun_0.45 \n[82] pkgconfig_2.0.3 \n```\n\n\n:::\n:::\n\n\n## References {.unnumbered}\n\n",
+ "markdown": "---\ntitle: \"Individual Participant Data Meta-analysis of clinical trials and real-world data\"\nauthors: \n - name: Pablo Verde\n affiliations:\n - ref: umcdusseldorf\n - name: Thomas Debray\n orcid: 0000-0002-1790-2719\n affiliations:\n - ref: smartdas\naffiliations:\n - id: smartdas\n name: Smart Data Analysis and Statistics B.V.\n city: Utrecht\n - id: umcdusseldorf\n name: Universitätsklinikum Düsseldorf\n city: Düsseldorf\nformat:\n html:\n toc: true\n number-sections: true\nexecute:\n cache: true\nbibliography: 'https://api.citedrive.com/bib/0d25b38b-db8f-43c4-b934-f4e2f3bd655a/references.bib?x=eyJpZCI6ICIwZDI1YjM4Yi1kYjhmLTQzYzQtYjkzNC1mNGUyZjNiZDY1NWEiLCAidXNlciI6ICIyNTA2IiwgInNpZ25hdHVyZSI6ICI0MGFkYjZhMzYyYWE5Y2U0MjQ2NWE2ZTQzNjlhMWY3NTk5MzhhNzUxZDNjYWIxNDlmYjM4NDgwOTYzMzY5YzFlIn0=/bibliography.bib'\n---\n\n\n\n\n## Introduction\n\n\n## Hierarchical Meta-Regression\nWe illustrate the implementation of hierarchical meta-regression using an example that involves the following data sources:\n\n* Aggregate data from 35 randomized trials investigating the efficacy of adjunctive treatments in managing diabetic foot problems compared with routine care\n* Individual participant data from a prospective cohort study investigating patient and limb survival in patients with diabetic foot ulcers \n\n### Aggregate data\nWe first retrieve the randomized evidence and summarize the treatment effect estimates using a random effects meta-analysis:\n\n\n::: {.cell}\n\n```{.r .cell-code}\nlibrary(dplyr)\nlibrary(jarbes)\nlibrary(metafor)\n\ndata(\"healing\")\n\naddat <- escalc(measure=\"OR\", ai=y_t, bi=n_t-y_t, ci=y_c, di=n_c-y_c, data=healing)\n\nresults.ADJ <- metagen(TE = yi, seTE = sqrt(vi),\n studlab = Study, data = addat,\n sm = \"OR\", \n prediction = TRUE)\n```\n:::\n\n\nThe corresponding forest plot is depicted below. The endpoint is healing without amputations within a period less than or equal to 1 year.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![](chapter_11_files/figure-html/unnamed-chunk-3-1.png){width=960}\n:::\n:::\n\n\nThe random effects meta-analysis yielded a pooled odds ratio of 1.89. However, substantial between-study heterogeneity was found, with $\\tau$ = 0.45.\n\n### Individual participant data\nSubsequently, we retrieve the individual participant data:\n\n\n::: {.cell}\n\n```{.r .cell-code}\ndata(\"healingipd\")\nIPD <- healingipd %>% dplyr::select(healing.without.amp, PAD, neuropathy,\n first.ever.lesion, no.continuous.care, \n male, diab.typ2, insulin, HOCHD, \n HOS, CRF, dialysis, DNOAP, smoking.ever, \n diabdur, wagner.class)\n```\n:::\n\n\nBriefly, these IPD were obtained from a prospective cohort study enrolling consecutive patients with diabetic foot ulcers (DFUs) and without previous major amputation in a single diabetes center between June 1998 and December 1999 [@morbach_long-term_2012]. The baseline characteristics of the study population are summarized below:\n\n\n::: {.cell}\n::: {.cell-output-display}\n\n```{=html}\n\n\n\n | \nHealing without amputation (N=165) | \nNo healing without amputation (N=95) | \nOverall (N=260) | \n
\n\n\n\nAge (years) | \n | \n | \n | \n
\n\nMean (SD) | \n69.1 (10.9) | \n68.5 (11.0) | \n68.9 (10.9) | \n
\n\nMedian [Min, Max] | \n70.0 [25.0, 90.0] | \n69.0 [36.0, 91.0] | \n70.0 [25.0, 91.0] | \n
\n\nDiabetes duration (years) | \n | \n | \n | \n
\n\nMean (SD) | \n15.9 (10.3) | \n15.9 (11.2) | \n15.9 (10.6) | \n
\n\nMedian [Min, Max] | \n14.0 [1.00, 53.0] | \n14.0 [0, 50.0] | \n14.0 [0, 53.0] | \n
\n\nSex | \n | \n | \n | \n
\n\nFemale | \n71 (43.0%) | \n35 (36.8%) | \n106 (40.8%) | \n
\n\nMale | \n94 (57.0%) | \n60 (63.2%) | \n154 (59.2%) | \n
\n\nEver smoker | \n | \n | \n | \n
\n\nYes | \n97 (58.8%) | \n57 (60.0%) | \n154 (59.2%) | \n
\n\nNo | \n68 (41.2%) | \n38 (40.0%) | \n106 (40.8%) | \n
\n\nDiabetes type 2 | \n | \n | \n | \n
\n\nYes | \n150 (90.9%) | \n79 (83.2%) | \n229 (88.1%) | \n
\n\nNo | \n15 (9.1%) | \n16 (16.8%) | \n31 (11.9%) | \n
\n\nPeripheral arterial disease | \n | \n | \n | \n
\n\nYes | \n82 (49.7%) | \n66 (69.5%) | \n148 (56.9%) | \n
\n\nNo | \n83 (50.3%) | \n29 (30.5%) | \n112 (43.1%) | \n
\n\nNeuropathy | \n | \n | \n | \n
\n\nYes | \n144 (87.3%) | \n80 (84.2%) | \n224 (86.2%) | \n
\n\nNo | \n21 (12.7%) | \n15 (15.8%) | \n36 (13.8%) | \n
\n\nFirst ever lesion | \n | \n | \n | \n
\n\nYes | \n70 (42.4%) | \n44 (46.3%) | \n114 (43.8%) | \n
\n\nNo | \n95 (57.6%) | \n51 (53.7%) | \n146 (56.2%) | \n
\n\nNo continuous care | \n | \n | \n | \n
\n\nYes | \n115 (69.7%) | \n62 (65.3%) | \n177 (68.1%) | \n
\n\nNo | \n50 (30.3%) | \n33 (34.7%) | \n83 (31.9%) | \n
\n\nInsulin dependent | \n | \n | \n | \n
\n\nYes | \n109 (66.1%) | \n65 (68.4%) | \n174 (66.9%) | \n
\n\nNo | \n56 (33.9%) | \n30 (31.6%) | \n86 (33.1%) | \n
\n\nHistory of coronary events (CHD) | \n | \n | \n | \n
\n\nYes | \n31 (18.8%) | \n21 (22.1%) | \n52 (20.0%) | \n
\n\nNo | \n134 (81.2%) | \n74 (77.9%) | \n208 (80.0%) | \n
\n\nHistory of stroke | \n | \n | \n | \n
\n\nYes | \n36 (21.8%) | \n19 (20.0%) | \n55 (21.2%) | \n
\n\nNo | \n129 (78.2%) | \n76 (80.0%) | \n205 (78.8%) | \n
\n\nCharcot foot syndrome | \n | \n | \n | \n
\n\nYes | \n28 (17.0%) | \n24 (25.3%) | \n52 (20.0%) | \n
\n\nNo | \n137 (83.0%) | \n71 (74.7%) | \n208 (80.0%) | \n
\n\nDialysis | \n | \n | \n | \n
\n\nYes | \n3 (1.8%) | \n6 (6.3%) | \n9 (3.5%) | \n
\n\nNo | \n162 (98.2%) | \n89 (93.7%) | \n251 (96.5%) | \n
\n\nDiabetic Neuropathic Osteoarthropathy (DNOAP) | \n | \n | \n | \n
\n\nYes | \n19 (11.5%) | \n10 (10.5%) | \n29 (11.2%) | \n
\n\nNo | \n146 (88.5%) | \n85 (89.5%) | \n231 (88.8%) | \n
\n\nWagner score | \n | \n | \n | \n
\n\n1-2 | \n115 (69.7%) | \n27 (28.4%) | \n142 (54.6%) | \n
\n\n3-4-5 | \n50 (30.3%) | \n68 (71.6%) | \n118 (45.4%) | \n
\n\n
\n
\n```\n\n:::\n:::\n\n\nAs depicted above, IPD are available from 260 patients. Some of these patients have similar characteristics to those enrolled in the randomized trials. However, other patients have comorbidities, where one or more risk factors prevent them to participate in the RCTs due to ethical reasons. For example,\n118 patients have severe ulcer lesions (Wagner score 3 to 5), and 77 patients suffer from severe ulcer lesions and peripheral arterial disease (PAD). The question is: Can we generalize the benefit of adjuvant therapies observed in the RCTs to the subgroups of patients encountered in clinical practice?\n\n### Hierarchical metaregression\nWe first investigate the event rate of patients receiving routine care:\n\n\n::: {.cell}\n::: {.cell-output-display}\n![](chapter_11_files/figure-html/unnamed-chunk-6-1.png){width=960}\n:::\n:::\n\n\nThe forest plot above indicates that the baseline risk in the observational study from Morbach et al. is much higher than most trials.\n\n\nWe fitted an HMR model to the available RWD and published AD: \n\n\n::: {.cell}\n\n```{.r .cell-code}\nset.seed(2022)\n\nAD <- healing %>% dplyr::select(yc = y_c, nc = n_c, \n yt = y_t, nt = n_t, Study = Study)\n\nmx2 <- hmr(data = AD, # Published aggregate data\n two.by.two = FALSE, # \n dataIPD = IPD, # Data frame of the IPD \n re = \"sm\", # Random effects model: \"sm\" scale mixtures \n link = \"logit\", # Link function of the random effects\n sd.mu.1 = 1, # Scale parameter for the prior of mu.1\n sd.mu.2 = 1, # Scale parameter for the prior of mu.2 \n sd.mu.phi = 1, # Scale parameter for the prior of mu.phi \n sigma.1.upper = 5, # Upper bound of the prior of sigma.1 \n sigma.2.upper = 5, # Upper bound of the prior of sigma.2\n sigma.beta.upper = 5, # Upper bound of the prior of sigma.beta\n sd.Fisher.rho = 1.25, # Scale parameter for the prior of rho\n df.estimate = TRUE, # If TRUE the degrees of freedom are estimated\n df.lower = 3, # Lower bound of the df's prior\n df.upper = 10, # Upper bound of the df's prior\n nr.chains = 2, # Number of MCMC chains\n nr.iterations = 10000, # Total number of iterations\n nr.adapt = 1000, # Number of iteration for burnin \n nr.thin = 1) # Thinning rate\n```\n:::\n\n\nWe start our analysis by visualizing the conflict of evidence between the different types of data and study types. The figure below depicts the posterior distribution of $\\mu_{\\phi}$, which is the mean bias of the IPD-NRS compared to the AD-RCTs control groups. With only one IPD-NRS, this parameter is partially\nidentifiable from the data. However, we can expect to learn about this bias parameter \nin a full Bayesian model.\n\n\n\n\n\n::: {.cell}\n\n```{.r .cell-code}\n# Pablo's calculations ...\nmu.phi <- mx2$BUGSoutput$sims.list$mu.phi\nmean(mu.phi)\n```\n\n::: {.cell-output .cell-output-stdout}\n\n```\n[1] 0.7868735\n```\n\n\n:::\n\n```{.r .cell-code}\nsd(mu.phi)\n```\n\n::: {.cell-output .cell-output-stdout}\n\n```\n[1] 1.091779\n```\n\n\n:::\n\n```{.r .cell-code}\n# Bias parameters: \n# mean sd 2.5% 25% 50% 75% 97.5% Rhat n.eff\n# mu.phi 0.79 1.09 -1.35 0.09 0.78 1.48 2.96 1 5000\n\n# Pr(mu.phi>0|Data) ...\n\nPr.mu.phi = sum(mu.phi > 0)/length(mu.phi)\nPr.mu.phi\n```\n\n::: {.cell-output .cell-output-stdout}\n\n```\n[1] 0.7733889\n```\n\n\n:::\n:::\n\n\nThe posterior distribution of $\\mu_{\\phi}$ has a mean of 0.79 and a 95% posterior interval of [-1.35, 2.96]. The posterior probability that $\\mu_{\\phi}$ is greater than zero is 77%, which indicates \nthat in average the IPD-NRS of this cohort present a better prognoses that the AD-RCTs control groups.\nThat means that taking the IPD-NRS results at face value would be misleading if we aim to combine them with a meta-analysis of AD-RCTs.\n\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Posterior sensitivity analysis of bias mean between the RCTs and the IPD-NRS.](chapter_11_files/figure-html/fig-hmr1-1.png){#fig-hmr1 width=1056}\n:::\n:::\n\n\n\n@fig-hmr2 presents the posterior distribution of the weights $w_{i}$ for each study included in the HMR. These posteriors are summarized using a forest plot, where posterior intervals substantially greater than one indicate outliers. One important aspect of the HMR is that those outliers are automatically down-weighted in the analysis.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Posterior distribution of the study weights, illustrated by the median and 95% credible intervals. Studies with posterior weights greater than 1.5, marked in red, are flagged as potential outliers.](chapter_11_files/figure-html/fig-hmr2-1.png){#fig-hmr2 width=576}\n:::\n:::\n\n\n@fig-hmr3 displays the results of the submodel corresponding to the\nIPD-NRS that received only medical routine care. The posteriors of the\nregression coefficients $\\beta_k$ ($k=1,\\dots, 15$) are summarized in a forest plot. This submodel\ndetects risk factors that can reduce the chance of getting healed. We see that\nthe group of patients with a Wagner score greater than 2 have substantially less\nchance of getting healed compared to the group with lower scores. This can also\nbe observed in the group of patients with PAD.\n\nInterestingly, these subgroups of patients that have lower chances of getting\nhealed are underrepresented in the RCTs populations. Therefore, by combining\nIPD-NRS with AD-RCT we can learn new insights about these patients that cannot\nbe learned neither from AD nor from IPD alone.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Posterior distribution of regression coefficients from the IPD-NRS analysis, illustrated by the mean and 95% credible intervals. The most relevant risk factors identified in this analysis were the Wagner classification (1-2 vs. 3-4-5) and the presence of peripheral arterial disease (PAD) (no vs. yes).](chapter_11_files/figure-html/fig-hmr3-1.png){#fig-hmr3 width=768}\n:::\n:::\n\n\nThe association between baseline healing risk without amputation within one year and the relative treatment effect is illustrated in @fig-hmr4. Results from the underlying HMR submodel are used to predict treatment effects across different patient subgroups, providing insights into how baseline risk impacts the effectiveness of the treatment. The posterior median and 95\\% credible intervals indicate that healthier patients (with a) are associated with a reduced treatment effect. In other words, healthier patients tend to derive less benefit from the adjunctive therapy compared to those with a higher baseline risk.\n\nThe model is centered at -0.565, corresponding to the posterior mean of $\\mu_1$, the RCTs' baseline risk. To the right of $\\mu_1$ we have the posterior mean of the IPD-NRS $\\mu_1 +\\mu_{\\phi}$, which has a posterior mean of 0.222. This shows an important bias captured by the introduction of $\\mu_{\\phi}$ in the model.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Summary results of generalizing relative treatment effects: The results of the RCTs are presented as a forest plot. The fitted hierarchical meta-regression model is depicted with solid lines representing the posterior median and 95% credible intervals.](chapter_11_files/figure-html/fig-hmr4-1.png){#fig-hmr4 width=768}\n:::\n:::\n\n\n@fig-hmr5 presents the posterior effectiveness contours of $(\\theta_{new,0}^l(B), \\delta_{new}^l(B))$ for the subgroups of patients not included in the RCTs and with low chances of getting healed. On the left panel we have the resulting contour for patients with PAD (i.e. $l=15$ and $\\beta_{15}$) and on the right panel for patients with Wagner score 3 and 4 (i.e. $l=1$ and $\\beta_1$).\n\nThe horizontal axis displays the uncertainty in the location of the baseline\nrisk $\\theta_{new,0}^l(B)$ of these subgroups. This uncertainty resulted from the\nposterior variability of $\\mu_1$, $\\mu_{\\phi}$, $\\beta_l$ and the amount of bias\ncorrection $B$. We can see that for both subgroups the posterior effectiveness\n$\\delta_{new}^l(B)$ is above the horizontal line of no effectiveness for the full range of $\\theta_{new,0}^l(B)$. If the clinical context is adequate, then these results indicate that these subgroup of patients may benefit from this new intervention.\n\n\n::: {.cell}\n::: {.cell-output-display}\n![Posterior contourns (50%, 75% and 95%) for the effectivenes for subgroups identified in the Hierarchical Meta-Regression analysis. Left panel: Subgroup of patients with PDA. Right panel: Subgroup of patients with Wagner score > 2.](chapter_11_files/figure-html/fig-hmr5-1.png){#fig-hmr5 width=768}\n:::\n:::\n\n\n\n\n## Version info {.unnumbered}\nThis chapter was rendered using the following version of R and its packages:\n\n\n::: {.cell}\n::: {.cell-output .cell-output-stdout}\n\n```\nR version 4.4.1 (2024-06-14 ucrt)\nPlatform: x86_64-w64-mingw32/x64\nRunning under: Windows 10 x64 (build 19045)\n\nMatrix products: default\n\n\nlocale:\n[1] LC_COLLATE=English_United Kingdom.utf8 \n[2] LC_CTYPE=English_United Kingdom.utf8 \n[3] LC_MONETARY=English_United Kingdom.utf8\n[4] LC_NUMERIC=C \n[5] LC_TIME=English_United Kingdom.utf8 \n\ntime zone: Europe/Paris\ntzcode source: internal\n\nattached base packages:\n[1] stats graphics grDevices utils datasets methods base \n\nother attached packages:\n [1] metafor_4.6-0 numDeriv_2016.8-1.1 Matrix_1.7-0 \n [4] baggr_0.7.8 Rcpp_1.0.13 meta_7.0-0 \n [7] metadat_1.2-0 table1_1.4.3 tableone_0.13.2 \n[10] dplyr_1.1.4 jarbes_2.2.1 GGally_2.2.1 \n[13] R2jags_0.8-5 rjags_4-15 mcmcplots_0.4.3 \n[16] coda_0.19-4.1 gridExtra_2.3 ggplot2_3.5.1 \n[19] kableExtra_1.4.0 \n\nloaded via a namespace (and not attached):\n [1] DBI_1.2.3 inline_0.3.19 testthat_3.2.1.1 \n [4] rlang_1.1.4 magrittr_2.0.3 matrixStats_1.4.1 \n [7] compiler_4.4.1 loo_2.8.0 systemfonts_1.1.0 \n [10] vctrs_0.6.5 stringr_1.5.1 pkgconfig_2.0.3 \n [13] crayon_1.5.3 fastmap_1.2.0 backports_1.5.0 \n [16] labeling_0.4.3 utf8_1.2.4 promises_1.3.0 \n [19] rmarkdown_2.28 tzdb_0.4.0 forestplot_3.1.3 \n [22] nloptr_2.1.1 R2WinBUGS_2.1-22.1 purrr_1.0.2 \n [25] xfun_0.45 jsonlite_1.8.8 later_1.3.2 \n [28] parallel_4.4.1 R6_2.5.1 StanHeaders_2.32.10\n [31] stringi_1.8.4 RColorBrewer_1.1-3 denstrip_1.5.4 \n [34] boot_1.3-31 brio_1.1.5 rstan_2.32.6 \n [37] knitr_1.48 readr_2.1.5 bayesplot_1.11.1 \n [40] httpuv_1.6.15 splines_4.4.1 tidyselect_1.2.1 \n [43] rstudioapi_0.16.0 abind_1.4-8 yaml_2.3.8 \n [46] codetools_0.2-20 miniUI_0.1.1.1 curl_5.2.1 \n [49] pkgbuild_1.4.4 lattice_0.22-6 tibble_3.2.1 \n [52] plyr_1.8.9 shiny_1.9.1 withr_3.0.1 \n [55] evaluate_0.24.0 gridGraphics_0.5-1 survival_3.7-0 \n [58] CompQuadForm_1.4.3 isoband_0.2.7 ggstats_0.6.0 \n [61] RcppParallel_5.1.9 survey_4.4-2 xml2_1.3.6 \n [64] pillar_1.9.0 stats4_4.4.1 checkmate_2.3.2 \n [67] generics_0.1.3 mathjaxr_1.6-0 hms_1.1.3 \n [70] rstantools_2.4.0 munsell_0.5.1 scales_1.3.0 \n [73] minqa_1.2.7 xtable_1.8-4 glue_1.7.0 \n [76] tools_4.4.1 lme4_1.1-35.5 fs_1.6.4 \n [79] grid_4.4.1 tidyr_1.3.1 mitools_2.4 \n [82] QuickJSR_1.3.1 colorspace_2.1-0 nlme_3.1-165 \n [85] sfsmisc_1.1-19 Formula_1.2-5 cli_3.6.3 \n [88] fansi_1.0.6 viridisLite_0.4.2 svglite_2.1.3 \n [91] V8_4.4.2 gtable_0.3.5 yulab.utils_0.1.7 \n [94] digest_0.6.36 ggrepel_0.9.6 ggplotify_0.1.2 \n [97] farver_2.1.2 htmlwidgets_1.6.4 htmltools_0.5.8.1 \n[100] lifecycle_1.0.4 mime_0.12 ggExtra_0.10.1 \n[103] MASS_7.3-61 \n```\n\n\n:::\n:::\n\n\n## References {.unnumbered}\n\n",
"supporting": [],
"filters": [
"rmarkdown/pagebreak.lua"
diff --git a/_freeze/chapter_11/figure-html/fig-hmr4-1.png b/_freeze/chapter_11/figure-html/fig-hmr4-1.png
index 5c726560141a74c440674f1b80590fd82c5aceb2..6d72aa8aee3606f77cbc8b27e73f30e6476777ea 100644
GIT binary patch
literal 26402
zcmeFa2|ShE_b9%f$2lB?V+duAGKNZ-IZ{a_Dx!o_3Ppw@GY=|FhDIgBDHU%DrOd-=
z5KU+>W;i7oGjmMGx%<$d-ut=t^Z)(t|MR*3d%xb(DbKTqwbxpEt+m&%-_sUbO?bE#
za{&N&OgFCE4gfp+CmWT(2LF0aTe}GWmXz0RZyN|J@$$8?<~Jx~`l*4x$=oIQHwU$z{Tx^0*&w^Y^S
zP48RH8TPcy=7UEb_vmNFWV&sOUl?KgKh0pz+
z{rVnH)U})RXXs4Vq_cNi_dpj|I22~v26{M`CN1k2Q`;Y?*I@j*E_m$4kt!gqL;yDj
zKJGx+6@0$)jb%&4T;MKyK)G0E@Z=P3;H{<%yNzw(#hZbKn?fw}Xhurm?|XJC^AH7b
zU|Xj$HeJ4!P0=}5xDdFae&a_1sCv4)k6bG4lw+I>zO*NFVEaRQzjL->zkEkae7@wh
zP%-n?puXK97BfT5jIYxkTwMt&G4C|r*@Cl=mB)^ZwHWOBTC&H#Iq0@^A&9uYtk-}5
z_6=Aj)>_9Elziwd?BCW!Vy39oIb0qj8yb?6#vE&s20I#7#l<{)*5y$sd9C>Ry_TSO
zJ>AgBr_Af2A-9KphRhPj;xg|yw<+%uNeKC_w#W05>cgUt+b#ZCHODzIAkD6i_dJ%X
zwW+T=;ShSYw9~}lPI=96NoLckWmTd6WA?|CMn5m7%cj1xpX8ZvuQf|rHWBybp4gby
z`;XS3tnsbLC7(+9hRcTOAD*`34eK4mQ8D+`dr?RjqcpIy?&Ymu%K*>kqjxvOS@~~!
z?7YR$ai&Z=xKo2sHo9?avPmO#BDej^m{`{QQDIms>QiA>MI&5xaniYCu(H?#?nYk%huGQ9l){oI7?1RAZw7d0YGDmk`3
zvQ$eZ$lq&c{c`N{PQhgW#S})gID^LQ2N(MI+_=?drTp*-(!o^h`4}GfB2sN(;*o!x
z-{`DQHkJ+Eie}cT+LcbA!_TI>9whDaX+}joSMR06{c^(I8c7tYD-BF!yZW}zAaV3?
z*Fa-q`-<#!-4Lz*2AWjb_E
z3P+mmlnv`wttcj!dt_)cXIdW|>oUbspA;I(@i|Epu6~7TgG*o0-V6fPr7?hZeYStWf>pi3G;K{v5`l4D$L>{Tf32S!!cfVpcQcW{|$>c5+EkffThdeWm0b-vt#N`PUNP`Yx;t_3r+(s?s1>
zu~;SaotbJ(WWuG4l7qLC!4=os_|IuqCHL(Vpx$=9yukR!V4u~nLC-INE^;Iw`#f;<
z-g1^yHA3=CDEYC%81vB6=m;P!9FFGo=<99T4QA3gby>Kn;G`)G8fq%3V6ipFC0HW&
zD{jD^TI_cMPlUv1gTy$I0M-(P0af5%av1L|7OV5B1
zsAgsNi=n3mt~rECd^S#@2w2{^xl-FIeEYVemf;~$;AYAnPnKZgX}HV(q;1)HnPT}-
z{4#|gZJ=ayKuPK1j-?6X{rO@y?4-+KII(~1D^thsGySnM5kprly!`SX
zflFuNQv^P42XwxW>C}FYS{%NZ1l59-{LGMnGR3S76-A*uJ&5X(QUr0pNjh%Wf`~rJ~
zXfgQKP=Ap{Fc+dAo$d4TCl?-4!oDa`+W9uKV6%<;#lCmiz!`mPilY~~bp9asN~1_#
z+UYGu|Db|AvyE6turFd`tBWHoql1ER+sKZ?eDAk^u`e`|_{r&DedX>(M+|_UB{;yZB
zP%YKTnP+WYVx+&=cA*H>O9peqWoggfdQZ{3U>km_ACL40;r-#f5>&x!b6V#RH_cBG
zThWjHJvr1hD-Rnj>#Rof&MW@XHd}0CId!~&`F^fheFIH_Ojq<&)&jA=_r(Fkb7AcA
z`5=5j$>NyI{3owavRu|5ZQi3~Rp}k`xA9{<)s1qc!p-iNo;<^iO(CL7=E)l+HPA3qfb7pPD*%Ko)>s9*m}tNyPJ)p-Rs;G;1sefOSIw{#x)7U`(VQi~T8gg%HtjkQ4DssuHL~T$aQy*|r{>l*YbG|Hgzm4sql7B^LQH
z83I|i;-{!S0~Q#8Pe#j9H~kQ2|Es6|cxt5ln!AzX4XEpON3dpFO~eQ+8NFHn^&FEW
zP#5FK3z{#lw)_bHK|s^Oz&Z5clWy$quZ^B5YrTPlxVax
z?a=!2kDIOLt?2GTi5+TIcoroONdtb%@Vh&`w-j40{EdPCGY8Sb4~%;xGxtT@hT58l
zt+?A<*mwHs{geo3Uz`(g}-!opE
zog)Wl5m#sLhq&$dUOR5xrRV)f6aNEKz=usT8HWfXqu`9s{L!;L1B{DNa*UtG>v>(u$jW&Lx0ibrS05q
zI59Q;if@KiD1@T`RDHwa!o_M=7DjzK+ItneKTi#;d2J}=;MI==1-6xCcE`xfVsmyc
zMdgB|sjRW;29CrT_SI;XWup|w)9@Ji~Vh`vC$j7T+YA;8yw%QqRk5h=+
zVV6w*&V#dXK4WXm%i>uDG;0|NOspdY-WN2|^Y5S;dht?T2<2voY2F{j6AsR_7@Rb}^;IgnEn0U^YuO&NQ{Yz-XC^_`ix%#h2kA2~?pv!A}Jf=Q{4
z2gtftbqzP=EZT8p8_9_L$SvC|+Ru1Eh-!LaT7mf`bj<$39^cZ=sXB*o#sY+ja-k`B
zWWOID&em+&^W!7#fx#!4q8wS3LiEg|S~xif4llB+O`n(C|Giy}Z>-r7>pxJ(l*V*+
zG!-0~l{Wm?>K0UCbb{w131aL-YZa0D=gUY;DG*eeO7}^qObjO9HJ^`ArYY@gW%6ce
zfU{uxu{=S~MV()Um=|yPEQWK}T$_i?L4DkaQvidRUDr3gd9r_%^(y-VYpRughyoT`
z$vT~@n{-hErE(rRc~B^)Kaze@$(-XmZkAz=z{Vhx20PMfT#uDJV7UwzU|0FfE
zBUpDlfq-h21W)C4<=-Ivkh8NZ_Z#2P(beh%&2a)%6i4PkvtC$(n%9`BG!U5w7%c<@
zR+T(Vo{8%%7A~0MJa)t;0^A!cm7<!sz?@9HEx6)#Lgj#fV
zOTby-+Q;FslR42+ix
zoHEVPB>pjC6@H{+gj&_GZ4L3N7uZT9@%-Ruvx@zmOr3suRiAx<_wq+f6CfZB9EsR!
zxUU+zDEyfAFYHYzQ92A7$#arwZdq-i=CtZAsaJD?ZFVb~)o-Mg{11MQVlsfLo2LLg
z@j%fuN9iE^KxeU>Hodnd+!8YP>p<
z5d6+#KDdI?n7PANz8S-rPXqZhB8L05He2A_|IIm6%E7dfgO6(i;QEP#_*k*02~4T6
z(1FNorvDi_+kh?C4}A?Mg!DLgp=jqDg14`s-#)E*l;SK<3|g_Y$@Ity&p#m0ShPn5
zZ4UYxypHms*8$cIGnP`nT;oD=Eefl
zf?80j!#a@|PnA|xa5dIz&2o7>rHc{>-4E)0nbCZFs}OO?-OR8JJ1uZo&}Lh0^IcQ(WF*zodG%QP}VS4Ru|Vz
zvNR!iIo7roloQCkv&zVui`M%#eUTarxnx^&+qbW*hKo7UqjQfytp&2m3TXXe$h=z-
zOA~Ab@m_iQvV5QA0W%{~J$ApQ=0Q6XsZ&-Ln+k{k&Ei08vTrw#N?T5SIvOu|RNpzK
zb$iam_2Hj1Ju1(74LgmQPrj}WTkYuJRUmp#9cngFQ9(ayZE2FH@8i
z0)cIWK5z{xM~Zgo^iR<)mwuD;q_T+TwXk{tQV%6l*D3+6ARBem0`me460l`MY(AI(
zzA=7M1*=`uF*(oUW6(;6tG{ClGN6d^Bq-3VpgwYPOynjYd9!+Vu%1oJPJ}v2W%1jx
z6=Djg=Z#tADT8G*kBoLJPo?jrT8GZ;?c5Hz^`|a2U>fk3cyz
zzsoomxrBRQH}m#$VUjnSAA#i*m{g@gk-G>bfiWmti#pJNrKURt!%uEqAT%MTj;bSw
zXsFz*N^c7m&{{M|${H^O1Zi;y1xs~Ta7#OSB5w|%X?mJv_gzl{BBl=|;Y0&z)kzUx
zjoc4|PoQp%m*A&5B}-kQn1XX`Y2iN}HWNwSI(X_v+G^ImW|<93umAUk@W4Qbh?4&c?uy20|kG4=ljN9Q3voxuIww6XBr
zWU#bp+ubnfoF&-R2!u$2VV=2xp&fMam_ybFT44msdL$VVskIQi<6P4Elf&biI#)=u
zbYdsRoKVXWd#YD&u1^dJRR8(-vaNE)&s+c~_ZGLNW6=*&m=phCd)Q>$I^;apu}Nc|
zKhOaUY>ugyFi;(b*lj(a{>qrNf(HXnjUb!B!*UdTBQ3>ky8qyX-tY2HawD5+4^ZAM
znD?Hg)VA_XCrBJ%3tPldJ#klnC7a+>p-}=MIRl$#*u%s1%B1Is(LfWA2T%{r*y`c!psvBXO(--|4X}k>Pz(MFTO}4Zi*2)8pokopXW-6;1Pa5@(Me>4ayxYKKChfv<@X
z4ynu#X26Wc>HEoSS!)1F;6Uc=h|h6hcYGs#x3be2J!P8A7`_mzz(Wu_N&q@qHC;Qn
z{mogg59w}p$@?rm)TFK(7aMer`}!^6@N(DZX3rgfWE^s!Nq4;KOSuf5%*BF?>X#op
z*el>AizHY_bD%oo8xy=(3}49=e8|ELreTk{3DgOhy_Wh8%akYkT(0;f=PQ-noxw
zd+TK=-f~&37PKi%`!ol(<3NKntG1rqy*WI?RPSM=EF#l!15J?z$-Y&C3@O9~1CuNI
z*YPHrLv)-|3fMEk7#>^n8t5*=viYXJA(JWD%sLU0znBRK*|H9lmp44X)ulqE)B8i_
zSW4A@{qj;?N^1I+0YS@=gxKui-QD~`UJi?D{H}qkKBm41v2ol1p*#a_ijwPa1D(Sw
z9r7GT2A`aD&`uWJ39c`uvu4;8Tin+<5?2h#x>iijccO8o_nRtLNqp
z7fHC8*NNe+MVERRwf>vNXP#`tXEii=x_;6^Js-@x_bxwJj<5Eb1=aFhJlJB8{UVx^
z@+N}8eIu7$Y&W>B5_lM*K($-#?bZZluZV$&IX>9^)m2>en3Vux^UR7=JCe`UDVHyv
zCv+J3nB#?`-5h&A{d}V88&jgtfyRy2vnv^OEPOo9wvv}@Duo>Hhgk^t1$c`DEoRx?
zfKj&rXP@;@h`@M{0FVuf8{wz6DS!W9l(|Y0!6ox<03nw41>+|#Ytyn^$B#2RRKwi|
z6mzccn|iVz3W3W|M!Fc4>Vty_ZZUZdQdG-6=w%~@y|?9k>o`O2e+cU1E5Ow)mC9)q$yP&U^uIflAid;0RYgdnxyy0u8r!(^9a
z9Iv=ARST#cf=WeI!l~H~Q5WzlwBeF=Pc3pp(|B<^N*8GOiD2iciCZke8~h{a7hA25
z^Q`R10aj~-FjwcpXS2grio|m4SV;QB0-JR@^hsDoh=aaaz}H;NS=1q2F~55rsIdQr2y;Q6p6&1`~%z>`7q-<$1n1eK75wORhLa}
zj`95#O(u|4;P#v3mfpWfOA*Nv!axBD^vnsUX4G$FLdRi$RZw5?CD>i8<5SczLPZzA
zRpU_`Xa}=S?b@70m1B1zTAo^wT(!oavJz23J6CbYp`zSyLm|j>2@cvKHpH7EgN0;M
zWw8fPb9YG+#^Q7r^)G$dl3rSTv_sQeKCsT9zJ&q4MC1DQd1<)3LDgON6sloK^su$S
z5rNL0KmorJ*{nQHYBNm&8?3~CEZY~yfgDd2O$L)KoWjUCJmkhq5WPt=Oq){_y&|g*
zZQaheecq@`X|OfV`Kr1_n;Nj$kZr6*xkhLf&ZMKFXJg;;WCawNe)tn9I)E
zwJn;*6(zIn-|uHSfT~XDn4>c;HXKZE|7k2)gH+R@pb#3fw^=Si*-b48Qq)z(gmAQwE>S$(SDH#jkfBpb6@y;o;VnWg
z593zMt@0c$XV&OvM{x3;0@QdMkwk;C0)|gqXAM#47Rt$!pr(du)_4K4gRiQjv36N@
z)H4coUSl!WqzFGS5M^{HqOw|mbZsomajap@zT>E;Z+#OVb@8nEgf);h?|~a46S@z@
zI4PfF->yuh!zI@%v!kM%chSz_$Bw$=Tr@e;v83Pwh77faz~$bG7Z;M)Be8gc^mOOL
zS_jX9d)qkgDqvY;4>T{1LT%TsH8)~5YJv6K4iTnG{lyifm`unt_*&weDjn*L_cWGdzTonIF
zryn-GSp@@pYn?K`xZlGtXl>`=J?x7gTM#H`2`Ej8a{q>acR4;Q$}45{bO$AxWw^Xm2G(VoGj;)9W_g*0PcH&qk%)9bf?I&VQZpSum@|vkS|I*N7
z^!K-2Z!nlS?BQj~t#Wvx2DTy`yxD-pSe&5eL63w7jtDcx}1(#mK~>7^S^zvhgOIlDj=vkxR}
zzRR#01~0yIQufNrZ2>h4XFHtXBtOtsZInVgO|F7{y}RxUJdzM;C;;8&i@7KVK$|dB
zxrQLwD*Zlzb=>v-iYgoqU5%tw;UCG^CNHSmjl9O)7u4lZ#dPc=PhI(lulDXdl8$;7
z%UOcb
zQx|3zCT)&%aIWj)$XaA<#*rxZ-Ok=6Jo$rBD#V&pvZS7K*
z{fM{E08;!iqnY1r6?>5s^%ZjcUc)AESWq|09{6;lHjHR6?YIQHzgGo3;|*PYu>5`b
z(@Jg%*_iVyXTh>#7X(R{eGrup6F?vf;kYP|YMedNM^IkGX4(DelVHtL5>OQD(Xqu8-1hy5vzXCK1+_6pqK(1m`K!Te%A#fSpq2d*M;zd3QN&Iq4rOrCI`d=J
z*u2uh#lGKxeh|clUFx0Uu8BP!GYR0)1DC}a;
zCpOPGRleP1>iYWogh`W2O?I87wOrofvp%-Wqmj3ORQL7$b^XXa=J}inXE859&H`B*
z55#Stb#(fI52DrDO{99&>BlU7RV(zC
zPK@>CunyYP%6sI2yIGamlG-F@LD#6`L~!vgU6tB`!lY**Tk}Btu>?Kb
zfTht=m%6Fz23a1F+F77P@f4j$0y@bY(R!j6q9UMimh6-CVes2aZpz@tmpORi(3SN=
zmBrV`Be5&!Dd@gpM=mw+LlQO>TdRNaN4)LCKOI@-vjg!ne^_q~ntN);&d;7Wc~HVu
z+Kgk)cL}$mWCM5DQ1L_1GM`llZcBEHElA1ZA|NWz_`(rr@227|?C*d7SnX-tN(8x=
z)V0$ni+6(>R_oe+JtoCX4yzp=jFJ*n#XgrK(ik2NABrTXOl7=&viaI2Xbamp);(iy
z6Q&M{do;=vnj+FEXh>?2S+((`-RA~ehTBzWeIwvd)trJ`KZ13Lxumg7*D@>cd}#tD7`MN6D{
z_T<)wgIl_t&`v$A{z>Y9nb&v&iJywHj@sVUFHe+pwr1PDEQ6jrjHgrRDFGMj5c{D%
z`I7_w>*{rNqhYMQdNyj#BQOGw*H30n*d=m3m4A|hS5K#t_D*s4HM5I|ankarqfdm}
zz$^Cgg6~-uF22Q84@%B~7knpzH(6{k;+8W|>qsnbE-p0a$sksEK=fj-yZ-S}!~ucRg&Sv2RXXjp+HG%Gdp_iB75T6Cmua?A1Y
zv5)MO2O`D365_hV@2E@t*)9R<4s&guSmBS(r!riP2vj_p;`&<{2rg&-!xxEwO>TrU$MY
zG>nn2#qP&c%6+s-{gR(-T~{LiZW(qs3ODA*H|tPY%=|~XuTSJtdyDIn%|@VDq{-)D
zcHjmLTSQz0O>L`Y&-A-8Z`LohjrDDJr5wyS5bxWS?@kUV;~V#IEm?Zui)t9
z!c_-8`Hl`~^X^l^+HQmB6mH6doacDu^Y=XA==itmo$QePd(VNPvop`s7?~Id_z;kv
zQg^LdLWndxrDKPa9y2lqWNTY0HbR!)eR}?bQp#tF}+fkQYerlTIezF;`w7`k`
z$4Hg#OGOOEr?Zs}#|MTN-YAbCl9)?#gI?0F^S|8+vg1Mb<`p8O;f=I0dag>-sPNI%
zLPsDB9wzV%l#Da*((t$w-Wg9h7(Ad5m0(CKj&oD?FxK4D;h67swiNZ
zBX9r;NQ1brq~MSahcS9cpmQ0=8~n%7y8`c=*hYz@in5^_vSF?=ZCn(AM?H+?Qh7^}
z{ds)IVFgg`y0)(Tyaw$tff{KiKs|Sn8o@=GP}tgW&c*220k{*ko*2qbX1Fn09Ci>%
zoq@|Mk#PbA-+BuE
z;^9Ob@6o4s-jjdps)IzDflk$&*pL6il6kl
z%N|F2nGuCs!_S6_E+j~6JxRk=2c52VlSRaH_6(c`de`3UA%)r3Zv-vRb{A}yg=YX_
z0(lSWBG&mhfkBqRCN;CW`yJHecj>C~O4j7o*L2)3Htb=1GH4yuLkE@wFr|8wD5Mp6
zY3)Vd4e#AnUX5-v0#7(|6fmpEgiy~N9Jc+Go+>p}!NWZiWptyt$|amd|G<4*b)G@`
z=z5@g%n);o*mbSwN64iKs5+!I@m{Z)guYxrEWSKegF=qFAea_wI=`ds(Q>SN4IM#*)wQ(M6GNd>+-ESKk)88Crb$(!N~qYYu3
z&o3l(DU&ayV|t57VDC}n02>FdZLXOIA}$rDIkb(FGSi}p;QCf*tXPKN#J9UmVlDmm
zImyq;7JI*HDdZeNwF<2!Ynf!`(~K74or7nSD8LCKMHZOX;;MHLWCchS>+w9v)un1~
z#as%04{8%1rgCrCkSzA}c&o&Qogf3ZgP+v)yox}*_}G}Fe?lF)uk3a<4eacZZpE
zC8O*GV`s_T&OAGHAH;7)|BfPd#q>@OUoDQd=-HqRW0|w=QcNdlt!NZ@$}YSn=mk5Y
zZ}9E)V8*Hhu<7;U&of~#ApCW5DDK04UtXw$7G5WiH8ap|zubJ#twGqk1tnbFee3h8
zHC?_Qt-XP`Lqe#+3Ss^&gK`5W*KYuy#5rgIpDMX1TdRyWd>+U16g{zc7U}lBzi-5i
z5^+kJwz7Iy%E>Tz?849i3VYdl*nv~;Ao8(GU~PZxcNo*erZ#H?Rct`7K4pNM0tcc7
zQ6=BIYx`u(07*S!`x5Lh^ob||k7MhA&zs}Ya#&KohyXSGz8^JI=~Us=`s2lQ;c+V^
z5m5}}XhYXXJ@n!{tcO0VO>#WAE$qHyUCP7ym~$(a9+I$<0(#4|Wstn74t8V&dXaWK
zde0#Pj2BPwRZ&JaKbpkg=^-GF+>siw0o0-{TAn+ILMdH5x65JbUsU8N%M5fa&?aEU
zq!9A}y;2nzgT!IEfF&90nXmP@DJC*lL*t
zx&op|{Rc&?+?rMM1{=ML!A1M?FGtk{Ngqtva9IZyhiCNjFVGPoqFLoTfq95;8;uqR
z)38gdkb0@n{;eM-7oeV{>Hhm(p68Mt5%BhhM@%;++gtWCpm&8n#{i>;L(y<%2jlc|
zF6krvt+97T7vX3@*01AUZsU>;TJkku9^wETkBm$Vj@M0)VRD+gV3
zG&E3ls*lO&QJXFqTOc3ixcp&{X-r3)n-#b(a_O7}2;LgnR|LAHv8B4pKxP-EUt;>v
zINciN@{`u>$5#iLPK@I`>pRejmz)89!i$^sq{J3L7r4$8JwzrPO8q4lP?uzz
zBm2>-KI3;eaWsQ<$3cPUMQ}(29MFRK>R%JTR9EF{3d1RL3RTU&q7Ki#wYVi
zTT7a+I?sGe8}id{-R?o3tiR8EtDXC4__n!it!qlv>$aAngI7)Z=>ubrri(ib-0MR4
zc88b*a!xYFM7y;I`kMnmY7d%MGPZ;zZ0|^8p9tg#dK|NbKpt-`#Cwi8WY`BE2<>lZ
z*PCK@!5S`CtD5JiQ%~iAuj7f}dPZFkODAZivWaDqd*ww+6
zhz6~m&4#!O*nVg7V<$kP{FQaU)+qtC5kco7(>!k0AX3xg%E=VBv5D)c2ZPxWt(Mgv
ztw{P^tNB-=^}XM(pf*xvQDD0YNBH4EB}!YH!}Qx$l??~nC)T_oP&a~zb8MZQ*2waP
ztSkB+T~{9x*N%U&OUdIWZ8`-i4b;X*`WL@Irtc2*bGah7QdaIjMv^bQT4`3d>LQWU
zaiQBhAR2d!{jh(f$LG&dR4oiIRLS(QJcsfJzvCM5)zZgL7Hi^S
zzbIoD*54`G^3s@v5_cSwSz%=!X(
zcvn^wMxClky@aoxk#MiW6DMKVW%5PsL~>E5A{OKyX}Y~njY!&5Ps;l673RpvNMonm
zc7f%spNXV}IE!G%%y-@kC(p6jEYArzML=%-jE}ahps>Te1nrgReoq1gW)_urxj_9S
zi|Mp=>k0&8SF6mt=vv7yqGWkBO+~1k9RqO=^mZ@ST5OdU1?-IBb$oa)iY(5FH1NL8*q4h2=
zibyJ{5$?TQgyh#bV
z?Q}TLG#6Z3I}TpwhAAold@>63^weP4L7B2^X^ti2=a5c6`dP&
zi4okCwbvH}aFd5uwHt}H2Bd~qm8_rhnHYHnU{H%
zxXF1sK$5!QX%E;MJ;LBv(pZTNH{@viZW%D*NSrKGo-`6OIH5bvH=Vr3~V$k=Zz(g@N277-nkWfA~3x
z`j!Pb7uA(R2=NOC;=Ne_MxwcKM1DuaThkCn<3v5P+EAO=o!*a4h*=|Vo!@Y;Jthq+
zfjAIv3@jjl@8=-7d5mDn8wTsD5J3S_=IxlH;?$EcDEEHmmGh5*x7dJ?9j51wK;r?v
z$m2?5!;zswxqF|9wn()TeDq=!F9kk>k*t#qq}Q_#1oGw{yf6W81KHY&q)4GKN?vjl
zkpbpJ3@XiA6S37}@0VcYV|zfeZp)O|`=7Z06arNcM+1l@2}nv~_h~+c*dkWqiq`D!
zTm-W6coYAdwl2(920PD+;K*R_D*>7YjR_Q0pKWJ$X)EVBYv3H3KWsEU8=i-Gpm4}X
zEA}CJ(+;HxZzW4L`1}mUThxE#-Fad_PxJ=Z`cRP+6JQ1vGprST0VxC(TkmKIJ-Ldw4lrcE9K8JX3o9=MN5iAD@chHd;j44LC_++K
zVj0YHIM8yCYWK9LmZ4ni6{3EuLU1YXd1iAK*>E(Y6!2ytEmUB9U%-TUfbM
z3-jmtV7}+DPYGcU*N?1M_{MlmD@Ugg#LP>b1vqY_+hRXZPqvhX6`|g2Kd^Z2%vL0U
z(Z2OXpy@+;w?}`|#M=)qTORX?a*AV{G-oqPd5P4t0e#MCztz+9zULkU!E^IpBy;Us
z&3IlxsPWbdj)dkDXgoU+74$E8lq9`Z6w+b2*Fc#3Ud1=gf7;%|Iz
zSWoq?u1tm*aX%5Eu7uu+2r)1m2px03aC|Gr86fm)F@9Arw^#1vD(g&pR@&w8jd=)E
zSArhAkU=#8q^^PdX-n6Rn##LxMz^6JcQZ3N31e>aXQg(eeodYFn(Syh_9l7Dx=gtz)C
zb6`l{mdwgZoFz6Zp(qZPU??mi)@Mj(11&
z>N6!r{$cj9&C3S&=BC}6iDD`Qtt7CLrr`c(VZN(Gar*w`*W^^wrxH)r#{OQUSt^ip
z_$PV9F7%e*y!L-$f(O=6&%;_!CGt^q(m$(88AA
zO0aB3weIsU%sAMOn$w6sn{Rl0Gs0tKlsC6@G4G?1C}yw1w2}o*8u?Mi0c(GPE*^aD
zsC3F(HLpX&ZswV7uv3Kb?3S@COdq_t``r`GYa4=`*z`?1K=4?uRLIOk>HguhrNTv*
zc88+6~sDngGE
z{#_2^dv+}t+;;F0FiMP8>D5_;jws<3|OOUv63nN+?_obED@~`at}^v#>+Ub>E51mi(g~ZyEdh8xvcL
z_7v{$)SVTXKdY2lmcOElQ-efkM+V!P@AZV{1yIChS?>38+m*4D?Y^NGn&oCRXe|T(
z>Q1b_|6Yd4F?FzxNp59|690z8fOAjx_hAPmJ|Ap;HFIwQfTb%0Y24)v%D-wNp@aWT
z^uxB9_GE1i+SyGhmwy#MLnC(U#jg`ZrG|l;rWsXEzVv*q3c)}6u14OJj>*r;yw^xNL;C(MBrTN3F1DM}GF}_>7!#i|$Ac`V=f#(y80nQ4F?bac;Fs=*qZUUb
ztA?zDw^;aZXtMP;$3I9I|J=Lt;OK9R$DLX$dgSCz2%Z{Ur2%X
zB)=|+`42PvhZ+7KGDG0Jy87S76^!oVVTgteXR-b-mA+bpS_~imeLc}*VQQ`>ktFfg
z1wkKefO(7dxOoCwn#^A}k}T$?L9_0NKwk7;*2E!YcyAJnZwXWRM1l@ie%BsL<^t+6
zDtH(xQ=1Ms2>{T@5~7xKlZ}Zm`15upyr){>3-(wU$!Z{iR@$~MkKNQeL>8{CuRS&TaNp)W3F
z`Bm8++kRhM0?&`(uPiNpT8^xnmo8K0U8BXicWG(byeqQc#^1)N^q(PE*MAkqG
zw4hM%)#T+3nJ{fLj_$az(4=gxqD?_^5+0<^Nih#%ZWr!}Jlbs>@1ihzM&G}KY_xlO
zyo*#xC?oZ9YhA&+ZCsRb+c@UO%aP($#>&xW7%N(n?8k#AU*~Gp7leBXJ=)Id>bfwU
z6I<388N;W9+VpQQQJf$7ZV#y#)O+;M{s9yd(`_oj3a$(2>ECe$o*GX!@f^H)Lt|&}
zC$WY&p?DW9a|464orjCXBvL?0@wTATuSyd$^uPK^?%tJIKiyN4>D*x4PzuA=hq64+
z6}=76&s5WQh7P@AE&KB5nP6tX1LefS
z*5)lYQn;SJ!zr)6b^7Fcn+Jgtbyo2%>U3V-nVC{69IbOhxdIk5td`oISTfi5%9#V`
z_{^vXvwtG@1ashJPQ*W76kKkzW{y|d*?aT-x@YE|^Y_t*zaJc2eOB;rNFLYj3>h~hoSR6imO1ml&C
zWtJGe71?Rn7*9Sde|cH=dgHQMF*!30!g|&{S=xPP7+&A~lo*%gmz$}l!^wnv%W8W}
z1bNoNKzRFY>SZr*dzLpc&gf5!)8t?8P#Lk_`;({o&Mc1^|FCXq+wwSwo)jR;=@u18
z>+VjsT;W(T$#3o1_5+s*3zeg0Zk`^ysRYiSrw$tqnIHNPZYg<0kMF(~1aVKXhsUkD
z`e3@$KWXyO;X4m)6yzz}^`PC)C*H;UF6!x>;wVJzEqtHgQ)X$=9(lJE<8WPQw=4tL!z^hO7`3}33zL^<2n`04>
zHpcB|k=e;Oz)a4K?SFrrZCZcf=!=?lv()>N=F%Q-9`C|;h{fpTig3%H;{AVnQ}tB=
zlb7Me&Mn=3-(h5>ozelXO?Jp&2)<~Tnlc@V$hA%$k1)Ix=liZ>)tX%;Qz<)_@xAh7C=J-q?zX2yWm+uL|7G
zQ0nRyIEVA3XR;MUtF+$zAaxcD`3&%ry0P3uwcRNSTPR>mnOiGSP30{RI)-mfqnPC#D;M_OUD~T%H);gntT^tH{oWoAUBU?__
z2#1Wd1YIm?8EYgX_m(>{a(TD9o?ip=e40_47eSXKpAB^@PL(Ba)82PS#boyCf0qY_
z-PhjFFqRI)84%uJIwQQ{Zmmxg@aiwEyzbTo6)Hs`CEekag}7
zF6riGzmP%d>5hOfVFG#6rQCsuZ_NekWwA#2Nlx<2-8|Iqd+o{>+a@>bA>vjgOT-^=
zXPuSC6t}fJ9Zu@*-FbxZA?{n?WY-IZdhr*>{`?r>DhGo=U&CT_|KOFXhstZuaNE@;
z`HD}gMawuO(QV^Iq;_{@z1pm|OZRl6G`8tYz`aA~RK8A`r<$x3R-QQ+pwTv=_FBKK
zh#*pOPer!BSGk8asp9+Q<_p)FHx9AO{U&a_)$Vbwd@;`(i)
z7frtla@cPMt6q-1C@SP=yC8f=
z$BL(ofqTBEE?%^O2Qt#*qabcLX%{NA2i>`WG&N^FkA)phF;}j=
zHWjwtRG9T8a-`YS8d(}xw<-DYnf&Rrs#dcz@^$IEgh?<)`Z!_qu?k+_R-8>IzJGPo
z5s}*^@n!GdvrR}H;8PR+qGWeF?BhShsf{6f&hPyQ>HMJasbQcjGb7QRI^jFio;xyO
zd@%L=n5jQUY^b5Cl;p0FUF{vVc#A!)M|W!Pw{y4|2Zki#zS+QxBzu)5j9`D2?psT+
z^%*HQr&K5FeUM9r>5AXJl#L!b#!0!7{qPF+${mR+!Vl=j_={$S9@p$qT^iV-bTQjI
zyUaveRWQBOxh5oj$+I*gu;Yu%eM7sE;`X;|`VGF>`VDR=d8ZcGVMX8OUt$HyHs3YD
zx;KvqP!CBh>7VX?P>b{2>lLJHxaq;GJWLxcS~n-yvwxo}QYa
zmICjRW8b@DU(w6UhVQdI>1M>co;I*XF6)XoL$aupUa3r8UR>CCS}ou?H?t~4F?Gkj
z?8!rxcPHOO_WAS(lZ-37cRQi0G14Ib@S?2$M}UGskIM{m`ow5lW~S4OOjI0B!k?Qm
zlXPI};KVcPDTsi5HCBnr88tTUxwqREPsH3#m$WW!NU7>8n(*GPo3>4_RSD~T?L3^t&iUi~j);&_=a*i!OT$tG*Jl&m|
zQ3DRUI%B9kn&ql`hcM##7iF)^8r-&>au`b4*mI_2Yl}W)PjY0lBT4GszhOSK4#%!q
zWi_GB&ZSQD8Sa!-Lv@%O-nV9ocE$KXM;RVqllENt0O17u8B;tYE9Li?qA@fB)lEGM
z-N=}6eOP~39}XiF=4oiolXC^Y2$bj@Jn$(8*r>fm>D2HL9Dg({2M`$1YRhgfBGK~o
z3?B>yAb_Qa*5WG_NR6kHaH|`FHKM;%wIPx8l$KGMnh}(92>*Y<1ly#hBcZ_1mx72>}~Ns+vgoPA_;99!ft$8R_4Tl
zJJFjU+nLtw@)A=l)p&eel{}pO1FQd0>sl@a5IHMf)=H1Fl_o=X^ReT>6*joyZ;Eu%
zBd=e~=GwJmhq&iptzZQLA$mg}Sg;+`iv&5sJBF|qSy37rA9&CD3Lm`kn^b~721~Z&
zVVSS~!VqCIhwq=XpFN_1e<=H2CO26BD9lm$r#dt7%WU22e%!}E`os^hJyQlRcSu9_Zi)6(+!FHY{86-OTXQY-?Vvex-oFN$Ur3vB0c7YNDc<4Wo#n>@ebZB
z*FvOwW5Ce`+RQ6T5*W>(1A{}
kpywl7{_jZl?RSaR&o`fzI8`Mtf=WLIxsK*22Q{r8aTo4EM;Ym
zj*bre7(V5G!u@hGFo-sAax!p&KeG&?qYa|r&st++js^G2-NGW$!rk4%9sW$Vh%C0Y
zw%)yaHyj-f>SO?5XD5SbZg1{aH2mWf4S!}iMb|oIWjSTRpPc>s_rp=RW4K4UM<%;R
z7P}`m1qXA&!^0!t=WIyyQySx%hd
zNZ5@#By7=?%xNm-G{Fzy4BYWx4-SVjr^9v|+y_C=)!V!PP&~o?i$pbl+Y0~*tXXNY
z`N-AIg0nHRZ30ub?Y5ie-9|}YX`fp(({0B6x7#HMT}31ZN1g57&xEC8XK38V3uRa1
zRcdyIq`rN#de^P3td`n=_YRjYY#CX$t*6AR?p2Fm+}p~QWgZD{?(}ywxrqS~vu7*Z
zWpExr09Y@D13+s75t7476ZlXP|8GB3jR1uN0H$jn2igH&PbfPLGZyCHGWS)^xwFn%
z?n2|V@20qD#aQZ+vmcwh1=l7XvJah!jsMm{3Tb+&nq-l`Blw1I5RMEP|&p^$-*+0x_H4$)p=-f)H>_9RR<2`;cjhMs$n8Q4zdsZxY#wO
zLUrERR|f@8gTBUTmA5yfg1H7;8DF9?6J@c$QDNC8Hi_pe)8IJL%|%Oin8
z`mf`%c>!iz=vUa(^U;a~1l)cdf-OSKe!Vi%=5k0~>izU*9ip?N(UBVC@ib23`1pQKo&B
zcg-Xh`&cg_^7~pK7y4st$my!j;Nz&b_;3&APqMNX*$>oPm
z3#)PJilmbk*C!cXscpL8b^|goZp9=O$337(KJvy$OO8!)%_X~^%yDT0CT>ELZ1FEZ
ze=+ozhIksmf6-Gmi01g49k;UC^*a-(v
zB2{S4c22!RDyw4&_6NstM(po@VEJmc|H~&{@Dsde2#9>?)b?;vY15Ua7vzf~mHYpk
zq7Kc;ktUbMfj_|dci1N!^n^OHXIM=1kIwOU`Zn_RsVv_M@l?c!akpMiDx#I!fW6@T
zqA4I|N&_!s&4Vf8r~prQqptm97-OI^RR9%$oLQGdrVRX-U;i$L|38yxvjmyX+&<;c
ztiQttu4=g?9{UB7x#D9;2c1d75>?Lq=0;J@~
zNkGybT4-`y#62Q|IXTe!(&Zx6B?PSZgvOogLL4Jk$pp9(G1;v)t}g#}KO)>H28vu_
zLWA!;32q`4k7_-E#@syQhu7acTBS}+u0{8cyY(N5=!kWyuTxgQ3
zNFhmYg_$w22`tmGMK~rviD96U1auD}0qhq`Q5Wr^Tl*(C4r)D1hAz-TpcYBRrcURv
z8Up$3c98J+4EVVIVB2;ct2+%UPvUz=Heb_nNujvg5s?@++G>r5Fk
zaXS53zcVLY9fNjs@4Rk@MNPR6?#^E~-I}oZ^`)?k2E569tb9k{*STIeAU_Xx505O|
zAjItI?>VGM*Vy!_aK{r366C!Pg;$^52$UC=b48E3GPGt~or0u)7qd_U`d{ev+kX7JQ~kT0>=64;`MwW&ayL$j2m*PE
z*qRBZNGiDQDMALdo+4NNcKRV;{~}@vB4n?U#_!G7a8To^8(co4rEfC4Ws`m0R-RJC
zr7ntRXm9e=LrJ7FabIjKrtl0B_C)Y$+LeZPnLJ$wawQWGCK>LiJB3$-0sgrJn(OIx
zAB1>X50}GT@5Rt<6|8`>_J<_<4{g71nsxBk$9XFx=q^xr{iHbc
zoSfmo>AnNT39p3AKw=2wZj*#|xTxrJ{yOASK{6ZJvUMIRD>{{)nnyX^hq}`A$a{~?
z#NLvj#QaIJsT>(j@X!hm>|ch=B+`qv0~rF<2ZnMcp_r5F{rLcay#LSUP46VP5iT*cOORaaOJjYzZlSd`BG5fe0^iJevdp4g
znqE#@T-$qLmBU{8$-Hi$D+WOG-85`$-RrCen=KD}-%*h`ll
zR@3;!YsJ)wBn!ltncK#gJ$K4kCA5_^Jfe?*=gOyedXk_GTJL&|J<{gohCJUfm0d|n
z(>KxF482DphVN1dbOc9YujY~V&LxaLncW&G`q;C=-v>r*EpP|b0%sC%Oynn+m|Z
zyS@sgH_bQs4aDj6U98}a66L~?VB_E=afatc)u&E*N{AU>(`lR*^o=%N<2Jal;djIY
zsZ^If7axkzL#FPDWT6o
zhm@JK5Q_@ZYVGa$ZLuryjOC)O2XTy{d%vl<)4BybJ4#y0r8&m=11XgrCQ~VMJP1T_
z;3jCQ+eaJ7D4^*8IcYqNE9_j~bqn;C^<`-d-3c_;!V&qoCO;H|Yx0%pVa>~_;C=;rsW7B_9>hmjSuX6{!3Gu855Q#4_>3|
z5}3ORx@p^+yJRU)K~X?Rz>v$>R$Hc90h#@XJqc4K0NzGhep!*eEr{>CAhW^U6j<;<
zv*gOG2?oCukDN&tBtkKg2|mUFs6>5*ow$7Z9nNzke`kH@)E&SfkB3T&$~aaWgFua(
zPE^0UQBsnG$^Mw;h;v==PiS$z4h>1lzE
z|IBZ8mX6SW{Fo0SpSzbgOPE#Y5$~;Uk5L*Kn&2zy8unw%eFm+W;9bk{YoXxl8~%?!
zhAKTpAPrpr5{RcUHfIv{d`wy(=)-=tK*=4%VkPjuc6kcUY07j-dBnbiniobVkbm*;
zmH6IPx4{ebn)u$gdQK(20`ex{s3->B-CYCp5<)>9@92bU;{01&75&FLt707UM*G_;
zJxYR!vWS8>D$741ri^459_34X`Bv}XM9swR^OHdv3GNUyu-9A?>26G@FWg>gG+2rG
z^F?{e%|rE}(kp_~l7r?jiv^aDutXE^jLcIvu112$vD4vGNPRZ)n5HsNs1=hlO%_?a
zPmsdY#=r|YOx>`SgbNdIe>{I&pM-f60Z%T$HMM!^;L?ptpEyIoT`B>1Qu=us?wa9X
zwuk$$*?~iikP2PBZ1U-
z3lK<(kQ}%+{~BMMya?(E5Ro4e-k5OJh>szE#-7Y`So(EO8S0JF1x56nv!d9SQ8n=}BziJT42PkNoO6G4*3N
zN{PsF15;3t_Hoy*GigBsd6s;BR~^WKzdxr4GoPe_9c#yVdchw{4av&2Zvb^ub@K|hmOkDVE
z4w%wSh!i#jtCkzVXfW>A$ESqIth8UNdn?2=X0yk~FZYc5wH!?4`?X86rqk2PVwd;F
z3`$N}JM@N#mnV#z7#WZ9xEeg5m*UJ7?Ob>uFmf=vA-HjTEW+5>|L{Ny^c(-g+1cJ8
zkWi4vF6BTUOV%}*dhl}t4}B5`$l@snG3E=mtzV$jaD=?k^3(n{E0tcq9L(Zqvmj;2AN}=ZtdjJ=IEbQ*vQlUuhz*DsSz%eNnM
z)}faOG82{E;_;P%WM)`kN3aGiW8yF!PkI|MkjHTi_^M66D@Z-6OTGj1O9ld-eXz4q
zfhsDG)?0C&52WdM)Ws;>9Q3?#avZMJ(;L6hakBTB7tV7Rs_*(~Cgeh^Dj;ab!9!W*
zb0`!(cAlCwp*V6+w&@%!Vn4FyBTj!25-|EHCD@>Wr%qF3{jKJibJ?@pz%k0_YGI21
z40Et(&P{Gi}+pC8wu2i{&(fBS?WTX@g&$j8_EXzQ*ksD1!r*ep&^Ye%dV*Y>Z0qp`RR1Yh7KTT>7i~inhPOa9=*wD)?w2rnMxF2zG3y;bm``LA
zy-Yh21=Fkqn%6Q3wVDM+hkaDB
z!WoO7Cip9UKWQZ)jOAjh@JP}rcox3ZYqE!K9TQfvdQ8kQKxWIs&qiIKuCeQN+?|D(
z>Ew2QWPz=lo_=<8A)U&@MF&{OW`qptY~#vci66=+<)~x202HFh>p`0h-pd#>f{Ji~
z)daQN1JeG%Mu8fKm8}n;tLsE7Z3|Q|F*(fxf}{b60vVG9XC$
z7OID1m^a?4SSsAgtl-&wl7_;q>_a`Ac5C`tzWJNyVP6D^WSIlV`bGon*^jw?*Hg!R
z5yf)=@qQc7w$DNE3~)t|Btb{PfZ(I7p0JwEs+xr(gVB(oeMQ$66OYRARIUaugyr*~
zv89jQw}R5s0V3C9AkT&N{k#*O8OVgf@8){`ArG4l)@xsn`k+OJS&kWC*Bw3;%*e;l
z&tQm;%xh%3Uf}cLTW75Sf+z7lIjHT6Gv$3GOQ4Qz_y4RxJqg)@GJyg&A1Vuij)cKV
zJM)0oi_mJF+KYX%WEwi<<25<+$5Xxsa^xe4^uiIJTLK~e&;a|@{nQS@@rsP1v=T@5
z+jNf^8d!nvC+0%ZQUXcEv}qr2hEvLXM|$Ff@t@=#skP2Syz9Nhkzx}dIAi7y+`Tgl
zsb{@=ibMtOe_TOM^H6_y&7t$Q8IB}?%-(Zy`B|RaKjn-Bj`ugGQjL}OFF4t8>ez?u
zJ>}|Hpm~_YFKE^ni+UgQ9ez6}OpSzDaxp-pYfRYg3$WJ-nm&B-_rc(NfP4=DK3vZN
zc4anVUpRDjrvP$X8Rss@{CX=x!B;L
zK&`1zv~DKSWS5IGLhm{A8PW&7QKn2JayzWd-xyP>5=z)3l2;G(bHutdCB+KXmSqJPa&LclW@nb4GZ3$W&G*9jy;v$JSxUA+F!
z7ABws67)i|V`#P4iYL(`=k83nBK(NMj`jBaQT3NTkM3+&*@=smz!6#J1Mf(QmrPPE
zzXC4EdvN&3CXaz|K6Lf?{XzmCK7HuMQ>Mk3`53A3ejeSMF
zq}p_0V488V=2K6^+e%IQCpa@^V&5$=NeTM+N5nAAHFTnp7Sc(}W&8xr=_D`;bblof
z5?jmFo%oU2L0cWepP)Id*R}_&p4^aL%G89wn=A%RHKd=lHEmjn)7qaP{Rrh@0M{)*eZf`JE7UNn1iSU@Pf&+0k!i3Avseuj
z#mL}P!h4yJWubR@gDI$BHIiREpmCadt#u<*`?%W%4Fuj&dKY1@>*suub}~S7Z0pGP
z1%Ll|*o1k0{+Btbm?Dwf;5=Xk?JC|9pW&mmRX?qSo+|n9-_YA8Lx0)R)0f|@({qBh
z_bH?`C+pj8<{?wax!8p1a7iq0Xp20NJPOAg?q$Rn{pnX0aMUGHR6xtj(Xeiz1CMqjT9qP|d`vgOLJSg9W
z(QV|>xH#}Q?lFRatxZekctVhLe4$l6K@37nc9jLCZcPr(4UgvVBhQV4>pSCrAZL!B
zdHM%;R>gp$QvCbF>xw^_Z);hLwBEG^G6If_p%MR<9;gS{hvo@UibXaz5(`z{RJkXd
zUq|2CBu)0=Ym&g6_A}4p8G!$56i;g19HEldU`%KTC4(XdO6`(Y7kl-JUdzJDLULL
z@ZL%wenaeGCy33{Rh9l!O4IJm}!V9D7iKYBklIaGxbc!q5>IXHv#%bO;!0+Xq
zOVt{VF@u+PSqzJ~yJf0AFAVnWtz19DiV^A}Vbiklh;PpPH(q_FG
z_F)P|GqFPR4rzMnIUJ60Y@M|ga{q;+BJErD8t8WIDZf+x{;=t#B{R0CZUQNDht8zW
z!Z8Asg_j|23G2eSPKWl{SXa@a9$!V%kBU#W`l3l)$F=f{r)#Z2I`gBB#R0i7iUb*E
z6LA$Le>6=DzuDLHl0&7uoiyr_F(ZV_?2V+IP?u}RLb$_i&j43w!SU6GA!%DjGw#%F
zzsmpp`wITUJ?QouTj)4^I;T@t6M0njY!z_G!KJJPDeB!7kf4Hq4Q7<*b7
zph%XyjAluCE6+GUa}~BhI&(t{#zJ3g$1|=)u{6+Dx~j_-RKK`QkYdEiP99)4=VyD+
zF6FT`=YX0>Y8-AX^2`#)_f%=*(LV9ku7URq}n2U?QA8NquBE|^ZPx}c3#-xfj?GR|hN^<);e
z)?M{7h54O9+X|(z{%vG?0pb%?mw$j-+^Dxvp})g2rVA5T`>v5&K?=YZ1X
z1HL(9>M7>@Y+}^t9Y>nmi1@ZbpW^bQ7Q>pLPD$sg?W^>H>mPj%Vv`CQb=0PdpPTQF
zxSjM6xG(uA3(Q>Fl4=Jcj1AI`*4CI$HiE%&j>Ke0^CbPT^$MpD_MM9h;5P-rDHO
zFib2D2s7E6JQTmFUl5`w8O=%GrfS)^klNwvojNk8gmV|Bjtog5ZicB6^itspnIh8<
zLCP>1!k~KKm2IUU-S>*_{KtbY*P;NuNO@n
zAdO`-eoj&bqa3T}dgJnERa@Vark{Q$T|HaBO)VsfNdDkavMe8KP{U5+Jcs874bQT&
z2Q#+z2112%AM|gX4gp&`Gw7uYzJ){2q>ep^Pw6C3S=+{JWRpL*vv+I+hP44FjQgJB
ztS9C;rNno{d2m=l0CB^m-!w&Q)uLlAi#{r)Ah@7)kqUgRlJtT8Mj<9+L&7uVZMh)p
z_{fPAT6RLuxIc`hT!CsN8%A8JV|=R~TXPhjMDEQp(+TNffl=87Sg}XR=`i{MWcGdP
zaUp8Y$DU`(OCN%bHcSU}xI(Sa*X%=w8gv?;QpVNY9`#4}U&^db_{!9gB$9hPN`#P4
zI0x3b^iaE?S6_SfK}*=Xt^P@H(F0U}*Dk2iA)QVK-m|CZy
zf|ZGPe!NxLKeQ2*mEc+*ZeBA{+Xs$(1SYhm#Ny9SCx>6#PZJN`(<~5@&GiM-m&{eg
zKBt776(Z6@SWOOdt=?;5n~a2AgmKibIDGH7a*uMSYiSTxn1d~0eDvc79A1xIRV!lm1C
zj0c0gqvD##^59Zyx*~m2$XIicYXG@q$@rFU8n@l>bq8DI;$CCNntn)z?O*MFo
z&QEUHx8kJ`)w%T3K!~Yy(Y2uo=jieE0`W$Ai_vP8HTd_!luV7fev4cjBj4k82OpFT
zT!j_-ZxwD^!KLKbR!k>FfzhZkpFk4YiXur%7ePxa>#%vcfQ}C`Q0N+Q{yjNc@G+DG
zD?)bTh*!e=h9*WSJtc^R(P8%R@m`6rQG7H!qtyzP`N}6j9^5cX3vNWnak#mU#%9T1
z#DAi$3
zpC+fTcA4_gfs1U>aF<7yK#npd0?!=RVK^$xm&(b59x~2X_OjHd`B+g<>@nbzV4E>u
z?LvE7c5#JTp9vDq*#~t@i0xB?%+)YSK(jlc-)TLvaL7Ex=W%vI&2+Nbs!9QiMI>y|
z#7QSmdS!+jzZu*BTgMVFF00GGJ+O2kma5#)_h6+e42I1TVkTvCeMh9FPaff+qm#Vs
zwFFu{^bwOFV*EUI0`q%<4_ctMJy9PC6lW*r2f)X1THUH@P6&a@JBV(uMTK#TMg3i^
zjWD<;%)EhtB_?T(GWGuFmoN1pIn_fa0n3Lw*zCCwvw_aYK!Ql{E42Ekb`c>kPPAar
z7_{!&C*f#?K2Ak4^HB8~K_)#`+!y7Zv6bA6N^Dl)DB}YCoZvxk^inpC!M>OW50kpN
z`y3J`MI>+G+Do&9nGDU5>+=^KLuH2+QVTR^d|AwpQk`fpo4|x&qHCq@pMwt43!V_^
z2OSGN^nzVB2BZLOdjZeh%wxFpv2G+za022v8d%)E>j{=HDZx|0-}1Yl{hY_f?=>!`
z;g-nIHO#x0-aDm^j2BHOFkLJyLBs%Ow^iLy_BvA$E{W4vo~)gZ)w93oJUMeF!slhhpHEk7H}SY#h@h+zDbhKs%eaCBWpUsl!9UPi7qlkdKC-!K(F)SHY5J%+u;X4
zrHhRg6!mZ-0?9U
zRk-hdT~Q#C4}tUmS;Ud$hZM=fS=$KImd|&4N(1DQtOZ|F?K(Kez#xnWtc0Fx?s;w-
z)f9K?MB4-3OON1)In(sNyvXHtxgbc1eX=Zi37X?W9~$&PNf=@793X~%n2*)vVyGA1
zAq@D{ITQe&H5;BLuxk<=SoI0487R@y6-GIij6Kzi9O6-5STGzW{AAQI(`+kL|6VE^tnw5^TaON{5u`p@
z(CR0{SOvzirxVGVKNx2Mdy;VmzKXe?5uR!S410_YBc02H=XD)fAx@w!m^LRYKcHtM
zSsj}jeEtYHf0yn>w}`Ja2eUT8u#KPrm$Q%$g>J`r7OdD|oa=#qyp-^1
zF7`r$nx8+!-Y878X%Ai)
z>Gw7TRI`8ro^e|pE3KEHYbe2sWMi+?zL{>9A9R<)$rSE{ap>)HfjSJ>L66>cq)UQM
zPqXsZ?j6^y6$r6IT>^FlA#O|e=YWrnmD1$#7Or)*0Q3KLHzludkScO>mA)eKM!7TV
z`@6CTG48QT804DSkw(yypS8*!oGsXDfIQ+P7Zmq6Lf`O#(}GObfCIv$QjqVMA22&k
z*m5<9<@A@ux0JgZ9t~9Nt!HZw*v*0YevL0SgRTycYU~jOs!qgh8jL!;YQIu;*<5VN
z*f9)(F(32)fUf?GtDc1^ZeF0Tf-GEZ++E9@jVH22-)#Wv@b^DFOHn;guW-GdJ@~~x
z29IYLcpY92O-x}@-~kNW^iVbfb_%ybVKlIDF~b{9S{O#Z@niX=B}{4BNzFN7wc8kZ
zcu$tym=PZi&nWasxe7k^)JX*Oj1jT#UAu*u$B|{HAa621G!3mDuN>lNS*->AH*3h6
zRr11PR*aLv)Pv%2ez^1@+43DkY_LlmQ5eY#3+>$Wq)cDm4Duk_*8cjbJD2^I9_RSeN73!u>3Wc_}xpDQVo@Ijp)}`F)6AC>v
zLFVa25|}#Tt$uYUV!zlaJ=GRo^&y{Z#A!iiPM~A*j`_+TXZFgF`(3e@P;-otrhnd#
zcyoMw`k;kOi%3#hjw8~h>Gx&kZhhqd8`76?EP!^Y%+`9(`-0n@1ElC%AjKeO1+gJ;ug5W5_w~XCJmE3
zjY}QKy22`E;LBaJyrr%M`PW1jFV5Uj|KY1N{kWLK=DtmiL|$@PsZ0m^P8xfNy>EwL
zK&v@4nvK#r-d+Tuiwu1;pt%GZJGLhL38NU8%rkAr88YfTR=pCae|Sr
zcFYsAT618R5IbOi`K99O0*7)?S+6eQbn-jVwE9L4|I?S?Vm`K2(j=jn1vdIk
zv|q~{)Vy3FL5>r`NBcy)HB5lfj(ItD^ZcY4tmZn-`|Frwt}>mGDnYKw#kS{PzN3ao
zZ?h3Uh%6Rkh#*L3V)3_-BzS^LulemiD95M}o^kb!233WGJ7Q@`n>)nVSE64G?D
z#+heT9FKt;XLg1B%Frs_9iWI&EpW_(>x#R@m&0&t=0nI_H+vz?7s!ae@r7dc8@OcRp2~!(T5-}XrlZy&fJi=b?{=6`
zP8|&L=pOA6pvKTk>LqPyAKco})*xA=^Uxubxt6Il;z39}Nd5?Oz2?us6kb}u)R>bS
zE`X1V!|C>9*B807s$EdqH%Q=f6Ko?|d}ve(1|$|)T7&)up3SfU2^i0)>3qOVN_hhF
zT-Q67!#FxI(G^)d>^Upkm{pP2BLGtnpNpd&jf-TgZzs>?&aiVHMxW``^hyNTTr^*p
zBnh)$o~KVE#}%xYP*Ti8!J)vt^)VqWXl1|4+k|T9EV-*Sk{79$244M>vp4uUOe8-v7;9kRX
zTy*e26l@IrQ=G16T=Q%ln022mfmwqw%lS}9>X5;K8xR=eh2-(AAb~U7+~3d54*B*C
z=lMxS)Y7YW29Ci>WyxIFx&W(OM5
z!S3eNB$toh91uZ^Pu>Aymgb+_4xbMpf=j+<QJbT)rb`go6`K^8Zg?2N4trEYs-cc3~d8z#gJ1Be8$Jm0ws(caPT
zlppSf*;EAPo^sTnH8~JkE_HReCQ$PV?;&K_wgiuU>O3{F13AA;*rjs77_}Cp79R$D
zUlwwe)bJQ7dSXV+x_%Z~eHA)S4ui&2js&(6+)BkM5k4UH(CxVW{vJ5k>@L*L`6y)c
zi?CfbXjDGi9LOP3SA<>W!NkAoSHZ<;WU#S8ASA)|Dm%lrzX#67hIhamcWY1`h~%OC
z6&WEu{D!^u-(Oc4hvs5i&r85uQ0iQ0FyS>zHG{S64hVSnnmchr$bPiD9VK)YFbGpR
zqQP!%OL*c&6@Dd1@t3h~z5h1MF#<@>wkP-$NxtTLlD*^O(7xdDZ|_Gr%ev}0b7x_O
zBrMRX+u!s8FWG~;MCiXx{NOnJ_PuffdteiEL9F9PJ_R;kxOr@Y0D9>j*k}x6&h>>H
zl4pXTPxDzIeW`9*pR3n+8hgP$FM9zLvJxU$;BMRoQ85CODQ)thdl@brQgQ^}Az@^;
zLUhFgSxAJq+3ryD#+r3)P7)u+A=h;sQUV5-&6H7rRja=AKUC$W1wm^
zF(|#vzryO4t5&B0*1hS71U+;IH)Y2y7G}r|Cc>&PH?AH7r)4R9D0GHKfh#0z&`!X9
z{aca-mg2C&U!%bk2(RZE(&G;U6eVToA!m=^V0z}Gx}{Br+WAfK4BfK=5CS#qb#S0p
zNx4&oo#uFNsUGLsI`}{KjsCSv)@pou;pNIzAW@$35!HV*F-_I76YG;IDa9JpK)nPw
z$qj)JecfvOBuc&-iGa2;!*QM`cjG*puGXy4hf5uROHGs}_o&Rn5pjsO=M}WNMt<>TV5s*w
zW;e{mYTiwtoV5k46HpYwOPWvGgs3VZvv8gVn*w!Wdg1sA$S3M_GLtBU65wgpw?N)}
z5X)iLmIzW)$DpW_zloky2z{lADe-cozc%5lYgteZ@Mi))aT67cM4<9P+Ij}3d*;n3
zx%v>Sn4D@cj$ttGYH!ztYj|%3h~-_#cMohPw?|kTfwB(WhbHwJU}_rmP?|j##-K(4
z^Df~u*ggW~ZL1bic~rI69A0*r(RDZbcz`hK2{T+UfLm`4Olbv$xtK<9`e98>MF?KJ
zSXgCq?gxTkqm8w0)ToOzR(+vYOPQVw=Yj70l~On+jA{*V`)>lLvSm`+gFu9T=;zTY
zzadV(bK4M~DNuqrY2pMj-5VcG3i2?qxN*qD&Tg{on;Hbf;WaLvV~iF
z7wyC`0CEQ9(GnT#7lZjni(%$92K=~r;dD`$QMF(iy?K9EOpsFiK@n1$s75lv%^M1SBy(X!U@BzQdI?=Y%2Jt6HcOJnqjtz^%PUN75)?^p^QjAvg+8O@PFC#R
z*d7|E{Pof6vM?z#=YD5vkhHJ06_3
zfu16B;*<0=-c=eBCPpMV(5=uIFB3F8c|%|qlvo53-RvW4cO)rj)9`NIeVmDr
zm@}KZ7s1q!e@r0v<40Qog3Iw6Sq&+t1m?IC`M9v43muJ5!&EhGuFgvaiX;qle27XH
z#HRL2>Rm`poHFnv=AZN6g6UjvV^-`y2*xF;jCRO)o}sq&~bC_g`OY%dIs_m5$>I(16_T=
z-dVN-#s(2RP#^o%!VxUMheN|6rOPIG-||X6E>ZkgPeEGj4AMgEg*40#da_p+S|AfU
zO7AceI?2sI8K_s_JK6A1?k6No19E)uLI7pS*eln-;E{@&&RZknpLL*hHMLdez;1WM
zp7qfpn4!0%9^KO}maA|L|neFf1>`Am!7u?{wSN+FRnu7zsZ+}-nxyZZAH!=U|
z;myPklEyITrK`SD1hO-3*#ZMH*s>_VpjB$h@I}3C)J?yL)OtF%f+pku6LfQL=3`+O
zADscKS1TFt0yDghH~|!1Af&(XC*Gld)sm0(%g?*me(TScX77$sGHSCu#Pg(Nr>^x*
z=#e2eL&@h8uG^n`es49rKhjswl9mOojc;f`nD}2?axb)v`rE9>VGLntlXLj0pX6jc
zoe#&o4x)iuH-%teG-7<}-f&BLzW<4l5LoeZ%2HNMqn|K8*mJsgH}(1*JBNXBqRNGk
z-M2fflaEj{@UR|fJzot+?PG
zP9Xz1o;c)(#ra&gRUj7gI6vgxsauEgVL0T>MOZF^%*RWg{$s?p>Gb9fU!S8P{^(rF
zl(CLORQwOdhBkK8TkchH7p{w#o8qUO$SqUBfTjv}a+2{X?j>=kMM0RM4>{<+2ZGa)
zL;_VFcF(^hJY}%|`oHa8jEFpke3_T!2REf!-C@cW{lEEJ<42coW=@)14iP4eoIL2@
z@IGRHVv)?{0`375*^VQQsf`dD_dE2^OU$C;J`z`{XY+
z{#PjBF#NKJdX@pKEmPN;msjy$B{>yE*pOEU9AOVBzqk5d|a$2pQ(9|`}nTJ6i
znPjZV{-3pN;RTrT9R)^0<3*y%uX25lu{R;0a`WS?cf!oD{xlD^Bm(G7caI!Um7&8c
zfCs)jtZ~fkzeI<=gh`VB)_I@O=gK31aZuP(od{ig_rhi
z;yZ$Il#rj>a5);d507{Cr)6rQJK$yDU;71Jd0bEJ^?t$MmVN^mRcCVziC@DC3NbTm
z%Jl*Y2BfCa!+L9YVNilRS0k+ICyl^o0#JGS*;nB|wvm6(5Ledxt?2uEjmD9`HIJuM
zoc`ad3avdh<30dN?c8?+{H+i@QC|G-iThh+{BPOw|NO%6-m{T_c`bF*N3HPS644FW
z^YUNH^ZsT1Q(rdlzsLIjvira69$r3#evH5D{x7@#&u9043*G-8Ds&H>aeuteV1QT9
zm%|7WnQE-thrZ%f9VSpu{qb^!Tv%+z9H{^Cx|e(S-jde@>e)XUKj3)^Xa_~8$NuFL
zn37~55K{XGhj}l={4$aeTrCZ=jiyNawOqG?j`8JL008mx+{(xK_5$e@A
zLfkj5jAOj`{jE0ea*TL+YybzAWgaO}Ksx_;g9R*;Azqd?((Z9Z^EWq3EcGe3=8*Uw
z7`rT@@h1}hoi|FrcZ>aDoW4?g@_)>JKY*v(M(_UhUZ%fqGMjw8(ckyE{iTKfBwCmy
z2-(#F90K2acwPb*9p>YyFD*CS^g4WBMX6;Xf%1@%BrCMVG|`m^+^8ECN!>fX>OBcm
zlM{E%+bqg`BSm`1sv_$oIqv%?D12v6U$@rfzKw#;|6`D|X>fe9Fbcq>c?19`H%I_r
zr$XVwIYNIt+z1O~alXMf!9gc*-3fzhC&IiJ4=M)-2eIFbUN7Ha=-BtZUTTeVvHSH>
zu^Bh+K9V0BKie-Fv3AA{)8WqdX?pC!47rqzpd(>mFe>kx^~-+G=*SmAncJ^j8!2hf
zOnCg}-G!Cr*UxhMb}p`(228yP9tnE5^n7kFj|AhfjuGX0{oq0I&*F!diwo2o==?IG
z!_jhF{zySYJgyNp+A145(fdGWOl&Oerq;@e5$Kt9+Ml2NR*oxOIvj6orNwtIul%-L
zVNdphWXH5ZV@IP$JP+E-O(ynUe&!wUDf{3=D3SC=bK}B9NKDowhA)%
zs_Vu%Bj@y#+K(uSJ*=&EPD|Ch75n;qi%0s6wKsMXmKf|gdHZ4Z1M#i2RPB{9Bh2yf
z9=44sP~^=%*dANab++(@sbcvv6S2ivQgr7uqcP47R#va0zSE4GiHubzMBZ*5)Xl0}
zx%cGl{wE*U=E-G|?_G|;Yf!AYnjV7!OL#LMqCP0CVvf4h4hi4XxtYx9;vwvSTmL#W
zQ*WXu;+?Wef{xQ?RJ)G&9m#w!d}4$%)abP&!_<3v-e(ol#TNq4Iw(t0PKrIX2XUJU
zU;XTT-=h)T2)+Z0@`OI~&`|h7$-PWy*GaKm2_V%b@vtza4g~$>{fGN>&Z@Vc
zxSA!~z#!Ut7ZWY^TJ4UDdGAg+IcH}f9s{#G_jiU34lbspztCTdvy;Co@|Nf?SF0E;
z@^-cY_vJ^2^5n(SPxIE&PLZWQ|L
zE%mV#Zsl6%-F7y?i&hG}mBLCc&~1#H(lT%3Vb%wvfSZXA#ItJ0nwpE2`UH=sEtZ?y
z0|WW|!X5D{2GO^H!PZ1wAqx9t%F>RiO(qA7E60U#)%(14a=ZN=MvGVJs2_^o8y%@K
zF_(GbSw*(UTb1RH3nSgjT_3GQ-f)H+lU?e!u{-s4@9el<G>P9jf`cy45sy)%f%YQTB_`(1b3rvR_wiFW7kdwyjWCU&)T4ehH_nEs+E)z_LOxj9TbVEIXcFwiwYjg&UkH;UutUoE-3ux
zNnL)L&zx3OBK8mIx+_C^n$jAe5oAm9o;Mc2=!pQWi@YVkU8E=G!pt+frKPY9j3=AF
z58XsPa~qEze)oB8x)B+yohh=VbkUcCILh_mIZtbKw&%}9DeR7Upz`bBZcP>N{B-*%
zJoR>?>$ona3-3!$V~ZpN67PJ^x!e5|B@Rf?_V0YPepC)QvhpLIe)*`Tmu>va=0i?G
zNcDiIa^&%4pK(7$p`3iiT^liXoHoO4|=c&^&&o+q>Yn#?~5f;fi+~`}`
zSKkyZZuZIGk$@;F-Sr?+dV>y1yi`&uRBFA7T{uf;h7U7VJMN&T>8wQ0r&Sh=6Jmo;
zWH-KC8@De&;piqCYsLJquU$=vn+fKT`)i*(d+WjK@K|g6|
z66JuWr>s|cBk5#@_Ksh=c=}uWg^!#3IzN72@TMZ6egC=D;3+b<^l|#Fb9#0-(xaHD
zcftD_?c(^Zh-Ep)oY_>ZrRKCbu|uR{Ph@oF8dlJX7?ZRRa2{vJr(bz)#q*ea-}bQ5
zs9mAH_hWY{?@3>lv%z^hYadDHsqf3tz|D1!gp0q5KRtu%Sac2*df-&N^egMK;Nmy;
zSA&(f!prp*Z$7#;yWPHM>+p}to5#^9i4NqjFY<>LdfZ-C8SKorN!C9`2-M@HYD&!4
zh{V?@SgWSW&DnRScjuM5Ju^A>fo{vzFX8Z+3hgv_5rU
z*WCR1D&_-q=+dui0G+~-t>K2b0UiRcmr5}L5uIXf9Gt#mRliU
z?7nq2uvg*1=0FRn+$pl1fV2>P<3D?gHcq|(VDE+M!Q3T
z`J*jwyb+AYwN#H?b^+1gk~Aw405>farKEKj#BD^}`gcU^Op4f?4g}~l0WEL^pk7Raz_9EUK$u_VX0-_bn;KhW*S9kLGbMH|DgbiQtVi(mR{-fOx)uOc
zyEdD+5>DxO1(0){KMMUi+&gg#xJ3C=W`OM1
z$E(*Q(x-L-sl#MlM>_M*0=gK4%4FKm2}Igt&}$$7CD9}RsrvwI9MsE=rfoe)mA46i
zc(pMKQ^OtZ1BeB{XQu7Ca%fhw0A%h1P##m)TgRKGVD|FYQLffbaeBfs1aKU0f`i9cS#dhGTvIN9i(ex9}Lft%GV0H!*>
z6L_v8w`wR~xJfM?Lni@n!IY29+=Q$C{5C
zCmW@OJhMU%Cwdl;seTp?(|+;6vw(80BYYm&S>k3{f}f>Gbm5zelFyM+UFe@L62P@2Yq)ccOu$HS~9a*1R(h|P*Wky
zXhmb_CIBu8#KxxW+&FQKjbpNQhNJWcADjP`GyK=bt6z0^8feCKl-VCI>GMLPHthG4
zFd6h32mn|Sb+7gb;%-#{AOK)hsN1Z*HEbg!0EGhq0IT746lZ`|PE9sK;*XpI0096i
z;cn80h$|n~a3bda7o)_21OUYW0RSt2Y_9*dOzpOoP5j|P03ZNh6-;{*aqrLcP8F;G
z0{{U40DxkF0000$p+Eot0H9DH0000`C=dVu04Nj)0000K3IqTE015>H0001m0s#O3
zfI@))002OtKmY&$pim$H002-Z5C8xGC=>_)000!q7m1ga#D5z^H~;_u07*qoM6N<$
Ef&{-P0ssI2
diff --git a/_freeze/chapter_11/figure-html/fig-hmr5-1.png b/_freeze/chapter_11/figure-html/fig-hmr5-1.png
new file mode 100644
index 0000000000000000000000000000000000000000..ada9814b43a5ab7fed523fa54f73acca87f7121d
GIT binary patch
literal 67755
zcmeGE2UJr**FTI-4hbc+&_R*jK|rwp7DACGQYMjVfsZfa=%PI(q{7699PtYycnud3hjj1LUKD4RTb_
z($eB^I7UWBW@ct~cJ|%7cXM%Z@$&Kt3JQvfii(MeNlHpeNlD4b$Q(FuKu%5$k+LOV
z>tQ1=A1!ZVBX5H|SI9?4%SR*6g9-`?ii(P=s;cVh>Zi6I4yR5
zy1E7i21Z6k78Vv(R#wQIklEPC14NvS4cL;m^@v9P*+e7H6*kd>HWkQ0va_>8WI1d-
zA{`?09U@yD@+Un!JiNWVgMxxWLqj8xksOc%`D8M(l@u)xqLC@uAjg)ttqE?uR3QJN
zE0E{G=t1Nl#l^)XCMITOW##4Nfo9sg*?P>gww4?H!DkAB?sctU!+F
zL8Rx_n8-8oARz(2gV9^!k*1URq{&v&
zB+>*K4UxUok3=G&t#$N~jiqu@H*y7FPw>`%RMdwBYXEqFhKk}@kK~!UjVA(Co?Pz1
zndeqr>F68B!~}H8IEuz?-8`?IFuEfclX6VM%R9E|%KlKsdpD+k2s9>yUN9=St7CFa
zMZQAm`Bdr~Nu93CXMOFkBT~nyb5yV0$9`3w8=Cv!fsOF;qwql0R!%Wcq9Hv&2T9`E
z+V`~Kp`&Xb)27r$e7;Sp4R9|{br!`?Y%a`qi^pvC?hXf3^zlFok-KBf_HPl`Yrksnr0~jqYFQab60;VH@jzv@pdWspxD2>
zAzdi0eWKu*b$;>$7WihUfrR@$n^YzHHfLwMpYi8y^y@tJ$z|sKazb=PD)swR=i%c&
z_aDLUQu=hXd8097euYBe0DWCGujyjiu|}WOBGuNDzO*S=AAI(iQS0W{;1grB!iT^9)IG94_L;f#p&6(ngc=Jl
zGAAuFlu@hM3Qw!n4t`Pm`SZwOolS;=L;dDI;zz31UOCNQG-snz^yrS?GsC?%_od4^|hQR%4IJ0T3`H1-(7NgUry6^
z$EHqY;lykep{`AfjX`dO-E%f+=iYtT6BalY-}u#|<)9aFl6oJL~g4EG?wOdOB`Duey+eRJAdHPbs*st5T_$VMjn{?Syc5
zY^_^Uw?o)IxkyFL%5^<>e_&R=TQD)4jDs5<=Zpg1UCsF#RHeL2T4Y^c3Km0OYaGpZr49KuUt
z&Q;(0zSyN>k4o>}7+2;bSDjSlUD28Lv6%_fb1jd~Za51ktr<43ar+EhP6^-^O3{1v
zC_7~YrE%cr!+p2~j!0DZt&^0X{@EV8n}^O2uH7|W$WQvZ@FHOi?^!}{SCo2H-T6BD
zHDJ3sbbZR)r*$Z8<$x^VqZ07>K=1qVJj7F7_?Dz%0%6`N9BX^ZR)L37g
z`(dGtVpnhKMhQSG?8HItk)!o1q}o5W;yxG|71@2mn
z?4u6RqU|itW8cTH<_(HevLn}I)@aIL`dA~HCBFu5H*S%AUr0A9mwJ;FJeq6own6xf
z%~xwtrO^zYdf^)}4*-#3`ghlwVn%-D&sq-v7jIl~oiRFiO|i*fmRko9*UXg}SiaPfWPy
z@2ZaeY$V;P_B|_BmZYJXdi_>aa9Q8&mB!DSgt4+m{>+GNb;bu31#ELOH}eVq;N6X|8C_V
zoM=K>9L>{*Vul4#GF7K--AaPGD;&F+1ei%Di$CJ
zb(1BFM;Ia=wbj80zGaQOg258+z==nxw?@F0%_1XsvTlvQVmpY8@J*uCK)u;igR<$M
z-!BQjb-gXkznz6q;!udE-`(0R;lDom|JM5}CdedT8Nm>2D84=6-|u@W;ojASR$
zfeEa*A&E4|EZf4B+q5D;oXm#c9;+{aZ24oheCD>*Y9
zhmx5$VOID=K~2TCTNmh)cMxUtHXrC9vUAxFJCw~@lgZ3$#p;7CVZ0ev11jq52;`}u
zudrZCo(enDmP$B0gfztgqM-p~ZC?L;j*Q2Z;CC^j3_|p}HMe{Uo6$s(LV&3J0vVS0
z#g8K)x1%(OU8eCDIvxL#B`rG)@gzcQ>g>oCZC0o?Oy3uXIgNt*&T2?r$^~JMG&<
zz$a9xIPG&5_^2B0jm4c7-_VsPO5{|?g3{@GRaU%;D!egMu{MagTq4e3c{Ae9ZlA4;
z&aR7siVqRoZx)AbucLje6usV}=f_O=7+Bm2;at(w8rGEw`;b9a7|Ldm(?IKzV5j|D
zZzO@&=&gy!AM6AZ8=7^hU=Tpj}oenQ6
zdUYK?o;5-AUBD><>iH>eddXyuBg0-WtxRpaxMw9wH%G`-@|?P+%v`Qob18+M$fM5-
z=2;t=_Un2l^-T(*JM4$5w#JpPIhay4d*~zOy9#dc&amn7(2s}aruQhu`pFHhUnwVf
zm2?$UT4q-sIMeLj`@;Fe;G>q!>E~jPExY*i?ullC0&44X->;PTmf8DZHl6yXl~I+=
z3=)UpRgEg@%+}NHbWW=*D3mVo?Rz3>zNlV!Q-FRvFXfJ1jnI10n+odn#CCDUn}Qct
zEYDm~-1S^6Z~flPGrGa2<2oVjuO705HrzRb%+abO5w0?Sb2F0pZt{S=_oWrzx+atT
zY7Dm;-q<}7=t!9gv6Oxys)TTZN0IphCK?GxdKsP*9L@xai0wnjgI^`r|2y`;R|BrI
z_$7o7^?BUrNC^lLocOVF&m44R%ckhOOnV$XIZ6yn;555)j*_^>99KKbe20_sgD@Hc
z-@ntJqI4%9^Jic>&^`OD;*7eBQ*0WJ83bvXUBhb7MSf%QhwyrBgW9}p^?E)^x`18G
zb68YBPWFWkB^I0zbUboXfI@AJN)biKnY@~fj3lDKD+2zXLH-|g{r|$c9$a*}Rwnzx
zcks7~mXeZYuDVdJVMYV-T4Z}ce>rI;;rS+glHCvLyuk$
z(|qr~WBEkgw&_2XFCyl-BNhvk1&jVzetOQN{U6JxiL0Ia<6aqep8Rny>9gOe|G3wG
z4C3^#-@ZE~NtBo2fAuRos(8mM|D)UAxm{)jJ)F7!>m(jXqI$g3MF8*r^!V2zfKKhg
z|L(z8sQ|)OX?_F1KWezl3@EfGdYMF&;`fgJmsK|~0-aM)P;}ETz&PzXa1O^h|IgwL
zQIsyRq`zQ8&J1q-Mv?T8t0;`z-wpgf)wE^b6IyTXxCB#V>*$zahJTffy{+)c!(SIj
zY@^vfi_rxlHY~tEscU=-9z_8Ee?VBGjAwP^SxfgV+t3B7f7Ok2LxlZLg(48Da`BsP
z`TFJWnE!^c$(FlkxllYU1RmD8W)1dZ9#HfWljUjMG!LgUmYzgnksl=O
z|8I!$g9x{O$DaQq4}?<>yb=E|{`^@FB=2}4S$n#xyNYSke}MrqEPUes!|E(=x-DH$!7l^bxG1cm|*0(mb`o+?(Hw11k~{i
z@M7O-eXAXhl1^=(xLfzflwOO2U5~BzW5MJ$g`g1utUCncc68f&|6=$bFzp~!dsf|3
z;Pv0ZjejkB&HOMHjuN1tz~b5eWjhv6`|S_ph?UfA0kjCPxX?N|SMu{8sp-IQXX7z{
zcXp|a?$LkYHtA|8rM3cX(Aw~4I|JB|9neKnYqlAPuFT7mL1=*w_;o~XX+O||S@d(Q!!w}GaqSJ*)D0+8%DXFBI
z9mCV{KRxle3aUr$(R#(krStH%?~A^E8J(zz&
z=eP+fWPnNbGXDNA>(`6jC_)%*LW^$b@D>z_`PKz5sorYF3?
z#3n>L(`kjY{q7eHYtBe+QPr)PkI#-gOPS#Kt{M4l8uuT(ZuT$INXE1?2?gj|UfT+B
zTLU2|>O@e|ip-et6=KsZKBW2&^`N$P@k0Go#4;uRve=`FcLWvo^fV9}!auEX{^g9zdF+j3|a5>kDgn(5})7-X&oC7Ws{t&L1;4Sp>ZBm#h
z-LGK2HJ1fw)WVmOx8A9;cgvQLmR6zADF5-Xta%{>BDPvc@*J)NtDS5Z!mh0|SiA#r
z!B1I8`0Vi9$r}D`U$%O^S!4u|fFE`BL;DAr&L@mDzaj~)`3;6hD~8zToq%mP$u$9!
zPXbMjA7<68j6M5HJ-VP_bZw&Sl)2ULIAvX~p?!~BmgozpW6P*0u<&jX;IplX3mZ`M
zxF;tQ_^wv(j7>feg>D9XkFZv#KQ6dw5?Q8bLNzsxz}z$|u&k8}PL+_nntp^FLmd2`
zdN-Pfm9PVmf3RSh4Hl_vHsg2qU!a1Ivd--HV01b3x1G~{ljgW+4n#X=#lXu1F|bR`
z5M(g@+vJZZBs-h?sUb5$9E=qH?Vbqgr~jJ?{tGO|k~Om&%8G`+qWFI>+MQeznGfu5yx$d(*9%
z6X_rUS=pK
zmpG~DXxm@gE(h_Hh^&8~3k!pAC+`GUo!XKnpjQ9%g|irAu-fdc6&5|pyTah+;4_Pl
zw#(L5e$agXS5nrhp(RRJw?aoDf*B(rT1xPQq6aoz2Teoh?Q3Nvbik%f*QIm|Iw_5C
zQn2|Mjcysl-dem&@KY702CIaQ^0zeG|}+B+~fd%@WU?>cCGnZd0O_YB
z&-(~T(oS;U&Ir4$M-_Ax34)qK$9ZE)8uFn?@cUi>uuaZN=uoRQ&pb$9Nmp7HIMBrM
zXM4IoEb5p1b|rmYaAroq6bQlgKN%M*)M7H7wIYPiKHUy8hQZ+H^L5f1XG6t_0{h^*rAo8=kWmA|6wm6>)jru3H$8`Mm=@N2Of(>a6
zLws`Wfq2Xa?twtoZ<8cPtqv-9nU5Y`mSMxeqH))fK{#jOF9<5YgKzJi`IVyaPRSjv
zj$rBhbBbH}e>*09FERu!lyDLah9?f$1uS|rZ$%-rkr_UGs7aI=_CPG*3JxedG6yE*
z9^csofhW`WN0E_jd_#o)tSsB`o>*h&ln>4PaoC#YAn$F%U$GV8)%6%O!BTfw^l&2Q
zhr!=jjPy5ZeEKGRE?)2tc(AovBh_;Oq;$h~4kUBJD*B5pO^^O*fO$6O#%ly`M2wab#3
zYI2h~$yF!{-BbEE(TWuTF>wTx`!9mQ6Mh}S?V7x)Re??imA?olkE3MT%MgmK5)l$+
z5FBldYs)(wwa+Zr@Sg-I16)vVmD^!oCp3lDt8