Conditional logit with non-reversible outcome

[Alessandr De Palma]

Dear Statalists,
I am trying to estimate the impact of temperatures on air conditioner adoption. I have an unbalanced panel from 2007 to 2022 at the household level. Unfortunately, I cannot show you any output given that I am working with sensitive data, and I have no access to internet when I work on them. My dependent variable is a binary variable = 1 when the family say that they have air conditioning in their dwelling. Now, once the AC variable = 1, it stays 1 for all the remaining years. So it is non-reversible. I am trying to run a conditional logistic regression with household fixed effects and linear time trends, but the likelihood does not converge: The output shows a big sequence of "(not concave)". Is it a problem with how my output have been constructed and how conditional logit with fixed effects works?
I hope that some of you can help me with this problem!
Thank you.

Alessandro.

[Carlo Lazzaro]

Alesandro:
the issue seems to rest on the evidence that your dependent variable shows limited/no within panel variation among data waves.
You may want to try a linear probability model via -xtreg,fe- (provided that -fe- estimator does not thrive on limited within-panel variation, too) and see what Stata gives you back.

[Alessandr De Palma]

Carlo Lazzaro If I run a LPM, I have significant results.

This is the transition of my variable from one state to another:

xttrans ac

| ac
ac | 0 1 | Total
-----------+----------------------+----------
0 | 83.37 16.63 | 100.00
1 | 0.00 100.00 | 100.00
-----------+----------------------+----------
Total | 12.12 87.88 | 100.00

[Carlo Lazzaro]

Alessandro:
provided that giving a true and fair representation of the data generating process outperforms any statistically significant result, as per FAQ, please post what you typed (I mean -xtreg,fe-) and what Stata gave you back. Thanks.

[Alessandr De Palma]

Sorry but I am having some problems to attach my output. I can only attach my code, which is:


xtreg ac mean20_cdd_30 cdd_30_shock_discounted_20 mean20_hdd_10 hdd_10_shock_discounted_20 $xvars rain sun snow c.syear , fe vce(cluster hid)

[Carlo Lazzaro]

Alessandro:
what's the issue with using CODE delimiters (#-shaped toggle) to share what you typed and what Stata gave you back?
Please note that your regression code cannot support any helpful reply.