Hello.
In many papers that use DiD (or staggered DiD or similar) work, I haven't seen one use new, powerful descriptive statistics tools like the dtable (or BenJan's dstat) package.
I am a retired professor and new to causal inference tools, so I work in close collaboration with AI programs like Claude and Gemini for coding assistance. During our DiD(csdid) analysis, I have noticed they take the lead and take the helm because I lack a strong grasp of the fundamentals. What I noticed is that , together, we go a long way , and then abruptly AI stops and starts complaining about how thin the cohort data is, etc. When I asked AI whether we could make a preliminary check on the thinness of the data or the shortness of post-treatment information by using the above descriptive tools , the AI replied that none could tell beforehand about the suitability of the data by merely using descriptive statistics tools. Since I am a newbie in the area , I cannot argue that these could help 100% but can't there cases whereby they might be useful ? Thank you for any reply.
Cem Payaslıoğlu.
Emeritus Professor of Economics
Stata 19.5 MP Single User 2 Core.
In many papers that use DiD (or staggered DiD or similar) work, I haven't seen one use new, powerful descriptive statistics tools like the dtable (or BenJan's dstat) package.
I am a retired professor and new to causal inference tools, so I work in close collaboration with AI programs like Claude and Gemini for coding assistance. During our DiD(csdid) analysis, I have noticed they take the lead and take the helm because I lack a strong grasp of the fundamentals. What I noticed is that , together, we go a long way , and then abruptly AI stops and starts complaining about how thin the cohort data is, etc. When I asked AI whether we could make a preliminary check on the thinness of the data or the shortness of post-treatment information by using the above descriptive tools , the AI replied that none could tell beforehand about the suitability of the data by merely using descriptive statistics tools. Since I am a newbie in the area , I cannot argue that these could help 100% but can't there cases whereby they might be useful ? Thank you for any reply.
Cem Payaslıoğlu.
Emeritus Professor of Economics
Stata 19.5 MP Single User 2 Core.
Comment