112 lines
3.8 KiB
R
112 lines
3.8 KiB
R
library(tidyverse)
|
|
library(dsl)
|
|
|
|
dsl_csv <-"111725_DSL_frame.csv"
|
|
dsl_df <- read.csv(dsl_csv, header = TRUE)
|
|
|
|
dsl_df <- dsl_df |>
|
|
dplyr::mutate(ttr_days = TTR / 24) |>
|
|
dplyr::mutate(task_resolution = dsl_score)
|
|
|
|
base_model <- dsl(
|
|
model = "logit",
|
|
formula = dsl_score ~ human_EP_prop_adac,
|
|
predicted_var = "human_EP_prop_adac",
|
|
prediction = "olmo_EP_prop_adac",
|
|
sample_prob = "sampling_prob",
|
|
data=dsl_df
|
|
)
|
|
summary(base_model)
|
|
|
|
case_model <- dsl(
|
|
model = "logit",
|
|
formula = dsl_score ~ human_EP_prop_adac + as.factor(source),
|
|
predicted_var = "human_EP_prop_adac",
|
|
prediction = "olmo_EP_prop_adac",
|
|
sample_prob = "sampling_prob",
|
|
data=dsl_df
|
|
)
|
|
summary(case_model)
|
|
|
|
logit_model <- dsl(
|
|
model = "logit",
|
|
formula = dsl_score ~ human_EP_prop_adac + human_TSOL_prop_adac + human_RK_prop_adac
|
|
+ week_index + as.factor(isAuthorWMF) + median_PC4_adac + median_PC3_adac + n_comments_before + as.factor(source) +
|
|
median_gerrit_reviewers,
|
|
predicted_var = c("human_EP_prop_adac", "human_TSOL_prop_adac", "human_RK_prop_adac"),
|
|
prediction = c("olmo_EP_prop_adac", "olmo_TSOL_prop_adac", "olmo_RK_prop_adac"),
|
|
sample_prob = "sampling_prob",
|
|
cluster="source",
|
|
cross_fit = 3,
|
|
sample_split = 20,
|
|
data=dsl_df
|
|
)
|
|
summary(logit_model)
|
|
#anova(dsl_df$olmo_RK_prop, dsl_df$median_gerrit_reviewers)
|
|
#chisq.test(table(dsl_df$isAuthorWMF, dsl_df$author_closer))
|
|
# https://cscu.cornell.edu/wp-content/uploads/clust.pdf
|
|
# https://statmodeling.stat.columbia.edu/2020/01/10/linear-or-logistic-regression-with-binary-outcomes/
|
|
# https://osf.io/preprints/psyarxiv/4gmbv_v1
|
|
felm_model <- dsl(
|
|
model = "felm",
|
|
formula = dsl_score ~ human_EP_prop_adac + human_TSOL_prop_adac + human_RK_prop_adac +
|
|
phase
|
|
+ median_PC4_adac + median_PC3_adac + n_comments_before + + isAuthorWMF,
|
|
predicted_var = c("human_EP_prop_adac", "human_TSOL_prop_adac", "human_RK_prop_adac"),
|
|
prediction = c("olmo_EP_prop_adac", "olmo_TSOL_prop_adac", "olmo_RK_prop_adac"),
|
|
sample_prob = "sampling_prob",
|
|
fixed_effect = "oneway",
|
|
index = c("source"),
|
|
cluster="source",
|
|
cross_fit = 3,
|
|
sample_split = 20,
|
|
data=felm_df
|
|
)
|
|
summary(felm_model)
|
|
#httpsfelm_model#https://github.com/naoki-egami/dsl/blob/537664a54163dda52ee277071fdfd9e8df2572a6/R/estimate_g.R#L39
|
|
|
|
|
|
dev_model <- dsl(
|
|
model = "logit",
|
|
formula = task_resolution ~ human_EP_prop_adac + human_TSOL_prop_adac + human_RK_prop_adac
|
|
+ median_PC4_adac + median_PC3_adac + n_comments_before
|
|
+ median_gerrit_reviewers + median_gerrit_loc_delta
|
|
+ week_index + as.factor(source) * as.factor(isAuthorWMF),
|
|
predicted_var = c("human_EP_prop_adac", "human_TSOL_prop_adac", "human_RK_prop_adac"),
|
|
prediction = c("olmo_EP_prop_adac", "olmo_TSOL_prop_adac", "olmo_RK_prop_adac"),
|
|
sample_prob = "sampling_prob",
|
|
cluster="source",
|
|
cross_fit = 3,
|
|
sample_split = 20,
|
|
data=dsl_df
|
|
)
|
|
summary(dev_model)
|
|
|
|
|
|
library(broom)
|
|
library(dplyr)
|
|
tidy.dsl <- function(x, conf.int = FALSE, conf.level = 0.95, exponentiate = FALSE, ...) {
|
|
res <- suppressMessages(dsl:::summary.dsl(object = x, ci = conf.level, ...))
|
|
terms <- row.names(res)
|
|
cols <- c("estimate" = "Estimate", "std.error" = "Std. Error", "p.value" = "p value")
|
|
if (conf.int) {
|
|
cols <- c(cols, "conf.low" = "CI Lower", "conf.high" = "CI Upper")
|
|
}
|
|
out <- as.list(res)[cols]
|
|
names(out) <- names(cols)
|
|
out <- as_tibble(as.data.frame(out))
|
|
out <- dplyr::bind_cols(term = terms, out)
|
|
if (exponentiate)
|
|
out <- broom:::exponentiate(out)
|
|
return(out)
|
|
}
|
|
coef_df <- tidy.dsl(dev_model)
|
|
ggplot(coef_df, aes(x = estimate, y = term)) +
|
|
geom_point(size = 1) +
|
|
geom_errorbar(aes(xmin = estimate - 1.96*std.error, xmax = estimate + 1.96 *std.error), height = 0.2) +
|
|
geom_vline(xintercept = 0, linetype = "dashed", color = "red") +
|
|
labs(title = "Fixed Effects Model Coefficients",
|
|
x = "Coefficient Estimate",
|
|
y = "Variable") +
|
|
theme_minimal()
|