add sth

analysis_regression.R

@@ -0,0 +1,25 @@
+# analysis based on mixed-effect regression model, where user login is the random effect
+
+library(lmerTest)
+
+data = read.csv(file="data_4_filtered_users.csv")
+data = data[!(data[['index']]==0),]
+data[['intervention']] = as.factor(data[['intervention']])
+
+m = lmer(
+    formula=log(commit_num+0.5)~
+        log(user_age+0.5) + 
+        log(history_commit_num+0.5) + 
+        log(history_pr_num+0.5) + 
+        log(history_issue_num+0.5) + 
+        log(history_issue_comment_num+0.5) + 
+        log(history_pr_review_num+0.5) + 
+        time + 
+        intervention + 
+        time_after_intervention + 
+        (1|login),
+    verbose=TRUE,
+    data=data
+)
+
+car::Anova(m, type=II)

collect_data.py

@@ -71,6 +71,7 @@ gap_df['pr_review_num'] = -1
 
 min_start_at = min(gap_df['start_at'])
 for row in gap_df.itertuples():
+
     login = getattr(row, 'login')
     # 1.
     cur.execute("select created_at from github_user where login=%s", (login,))
@@ -80,7 +81,11 @@ for row in gap_df.itertuples():
 
     # 2. 
     cur.execute("select sum(contribution_count) as history_commit_num from github_user_commits_per_day where date<%s and login=%s", (start_at,login))
-    history_commit_num = int(cur.fetchone()['history_commit_num'])
+    history_commit_num = cur.fetchone()['history_commit_num']
+    if history_commit_num is None:
+        history_commit_num = 0
+    else:
+        history_commit_num = int(history_commit_num)
 
     # 3. 
     cur.execute("select count(*) as history_pr_num from github_user_pr where created_at<%s and login=%s", (start_at, login))
@@ -126,7 +131,11 @@ for row in gap_df.itertuples():
     # 10.
     end_at = getattr(row, 'end_at')
     cur.execute("select sum(contribution_count) as commit_num from github_user_commits_per_day where date>%s and date<=%s and login=%s", (start_at, end_at, login))
-    commit_num = int(cur.fetchone()['commit_num'])
+    commit_num = cur.fetchone()['commit_num']
+    if commit_num is None:
+        commit_num = 0
+    else:
+        commit_num = int(commit_num)
 
     # 11.
     cur.execute("select count(*) as pr_num from github_user_pr where created_at>%s and created_at<=%s and login=%s", (start_at, end_at, login))