updating for new bivariate plots

analysis_data/100725_bivariate_plots.R

@@ -15,7 +15,7 @@ library(ggdist)
 
 ggplot(main_df, aes(x = week_index, y = n_comments, fill = isAuthorWMF)) +
   facet_grid(~source) +
-  geom_dots(side = "both", layout = "hex", stackratio = 0.92) +
+  geom_dots(side=  "both", layout = "hex", stackratio = 0.92) +
   scale_fill_viridis_d() + 
   xlim(-130, 15) + 
   theme_minimal() +

analysis_data/style_dict_variables.R

@@ -7,6 +7,19 @@ library(purrr)
 unified_csv <-"~/analysis_data/102725_unified.csv"
 unified_df <- read.csv(unified_csv, header = TRUE) 
 
+unified_df <- unified_df |> mutate(has_update = !is.na(olmo_sentence_labels) &
+                                     grepl("(ACTION ON ISSUE|TASK PROGRESS)", as.character(olmo_sentence_labels),
+                                           ignore.case = TRUE),
+                                   has_update = ifelse(
+                                     has_update, 1, 0
+                                   ))
+
+model <- glm(has_update ~ PC1 + PC2 + PC3 + PC4 + modal_verbs,
+             family = binomial(link = "logit"),
+             data = unified_df)
+summary(model)
+
+
 
 unified_df |>
   ggplot(

dsl/dsl.R

@@ -0,0 +1,56 @@
+library(tidyverse)
+library(dsl)
+
+dsl_csv <-"~/dsl/102725_DSL_df_adac.csv"
+dsl_df <- read.csv(dsl_csv, header = TRUE) 
+
+
+power_model <- power_dsl(
+  labeled_size = c(100, 200, 300, 600, 1000),
+  model = "logit", 
+  formula = dsl_score ~ human_SOL_prop_adac + 
+    median_gerrit_loc_delta + median_gerrit_reviewers + 
+    as.factor(isAuthorWMF) + 
+    as.factor(source) +
+    median_PC3_adac +
+    week_index,
+  predicted_var = "human_SOL_prop_adac",
+  prediction = "olmo_SOL_prop_adac",
+  sample_prob = "sampling_prob",
+  data=dsl_df
+)
+summary(power_model)
+plot(power_model, coef_name = "human_SOL_prop_adac")
+
+dsl_df <- dsl_df |>
+  filter(source=="c1")
+
+trial_model <- dsl(
+  model = "logit", 
+  formula = dsl_score ~ human_BI_prop_adac + 
+    median_gerrit_loc_delta + median_gerrit_reviewers + 
+    as.factor(isAuthorWMF) + 
+    as.factor(author_closer) + 
+    median_PC4_adac +
+    week_index,
+  predicted_var = "human_BI_prop_adac",
+  prediction = "olmo_BI_prop_adac",
+  sample_prob = "sampling_prob",
+  data=dsl_df
+)
+summary(trial_model)
+
+style_model <- dsl(
+  model = "lm", 
+  formula = human_BE_prop ~ 
+    median_PC1 + median_PC4 + 
+    as.factor(isAuthorWMF) + 
+    as.factor(author_closer) + 
+    median_PC3 +
+    week_index,
+  predicted_var = "human_BE_prop",
+  prediction = "olmo_BE_prop",
+  sample_prob = "sampling_prob",
+  data=dsl_df
+)
+summary(style_model)

dsl/final_bivariate.R

@@ -0,0 +1,45 @@
+library(tidyverse)
+#library(dsl)
+library(dplyr)
+dsl_csv <-"~/dsl/102725_DSL_df_adac.csv"
+dsl_df <- read.csv(dsl_csv, header = TRUE) 
+
+outcome_summary <- dsl_df |>
+  group_by(source, isAuthorWMF)|>
+  summarise(
+    total_sum = sum(!is.na(resolution_outcome)),
+    count_resolution_outcome = sum(resolution_outcome),
+    success_prop = count_resolution_outcome / total_sum,
+    median_ttr_days = median(TTR, na.rm = TRUE) / 24
+    )
+
+
+library(ggplot2)
+library(ggdist)
+
+
+signed_power <- function(x, p) {
+  sign(x) * abs(x) ^ p
+}
+
+signed_log <- function(x) sign(x) * log1p(abs(x))  
+dsl_df <- dsl_df |>
+  mutate(
+    sp_med_pc3_adac = signed_power(median_PC3_adac, 0.2),
+    sp_med_pc4_adac = signed_power(median_PC4_adac, 0.2),
+    sl_med_pc4_adac = signed_log(median_PC4_adac),
+    sl_med_pc3_adac = signed_log(median_PC3_adac)
+  )
+
+
+ggplot(dsl_df, aes(
+  y= log1p(TTR/24),
+  x=sl_med_pc4_adac,
+  shape=isAuthorWMF,
+  color=isAuthorWMF
+                   )) +
+  facet_grid(~source) +
+  theme_minimal() + 
+  geom_smooth(method="loess", span=0.5) + 
+  geom_point() + 
+  scale_color_viridis_d() 

dsl/survival.R

@@ -3,8 +3,6 @@ library(tidyverse)
 dsl_csv <-"~/dsl/102725_DSL_df_adac.csv"
 dsl_df <- read.csv(dsl_csv, header = TRUE) 
 #https://stats.oarc.ucla.edu/wp-content/uploads/2025/02/survival_r_full.html
-dsl_df <- dsl_df |>
-  filter(source == "c1")
 
 library(survival)
 library(broom)
@@ -12,8 +10,8 @@ dsl_df$ttr_weeks <- dsl_df$TTR / 168
 trial.survival <- Surv(dsl_df$ttr_weeks)
 trial.model <- coxph(trial.survival ~ isAuthorWMF + 
                        median_PC3_adac + week_index + 
-                       median_gerrit_loc_delta + median_gerrit_reviewers +
-                       olmo_BI_prop_adac, data=dsl_df)
+                       median_gerrit_loc_delta + median_gerrit_reviewers + source +
+                     phase + author_closer, data=dsl_df)
 summary(trial.model)
 trial.tab <- tidy(trial.model,  exponentiate=T, conf.int=T)
 
@@ -26,3 +24,5 @@ ggplot(trial.tab,
 
 surv.at.means <- survfit(trial.model)
 plot(surv.at.means, xlab="weeks", ylab="survival probability")
+#https://hbiostat.org/stat/binarysurv
+plot(surv.at.means, xlab = "weeks", ylab = "survival probability", xlim = c(0, 26))

mgaughan-rstudio-server_30651103.out

@@ -0,0 +1,17 @@
+1. SSH tunnel from your workstation using the following command:
+
+   ssh -N -L 8787:n3439:51687 mjilg@klone.hyak.uw.edu
+
+   and point your web browser to http://localhost:8787
+
+2. log in to RStudio Server using the following credentials:
+
+   user: mjilg
+   password: JZoMumQwFbIc7EUcaGbK
+
+When done using RStudio Server, terminate the job by:
+
+1. Exit the RStudio Session ("power" button in the top right corner of the RStudio window)
+2. Issue the following command on the login node:
+
+      scancel -f 30651103