xtlogit vs logistic model

[Adamu Idris]

Dear respected Members,
Pls could anyone help and offer me an explanation with regards to the below question(s);
Why is it that each time I run a xtlogit and logistic regression the number of variables that may be significant differs? for instance, i have six IVs, one moderator variable, four interaction terms and three control variables. When I use xtlogit, the moderator variable may not be significant, but when I use logistic it will be significant along with other variables that were not significant when xtlogit is used and vice versa.

Thank you
Adamu Idris

[Carlo Lazzaro]

Adamu:
-you are comparing two different regression models: no wonder that their results do not overlap;
- what's the rationale supporting the interchangeability of -logit- and -xtlogit- when it comes to panel data analysis?;
- did you cluster the standard errors on -panelid- when you ran -logit- with panel data?

That said: as per FAQ, please note that your chances of getting helpful replies are conditional on posting what you typed and what Stata gave you back (the described approach worths more than tons of words trying to explain qualitatively a quantitative matter). Thanks.

[Adamu Idris]

Sorry pls, I forgot to mention that the moderator variable is a categorical variable with "0 and 1" while all other variables are continuous.

Adamu Idris

[Carlo Lazzaro]

Adamu:
thanks for providing that further detail.
However, my previous comments still hold.

[Adamu Idris]

Thank you Carlo
The rationale behind running xtlogit and logistic is to be able carry out some post estimations such as GOF, LR test and the like. But pls, correct me if am doing the wrong thing. That is which of them should i stick to and the likely post estimation to be done after each.

Thank you so much for having time to respond to my questions.

Adamu Idris

[Carlo Lazzaro]

Adamu:
I find your rationale a bit weak to defend.
If you have a panel data set with a binary dependent variable, -logit- outperforms -xtlogit- only when the LR test appearing as a footnote under the -xtlogit- outome table fails to reach staistical significance. If that were the case, you should cluster your stndard errors on -panelid-, otherwise, under -logit-, Stata will ignore the panel structure of your datset.
Otherwise, you should go -xtlogit- .
Eventually, -xtlogit postestimation- offers many post estimation tools.

[Adamu Idris]

Carlo
Thank you once again for having you at this crucial time of my study. If I understood your argument given my issue at hand (panel data set) I should ignore logit and focus on xtlogit right? If yes, Pls Carlo assist me with some post-estimation tools for xtlogit that suites my context (combination of categorical data as moderator variable and continuous variables as my independent variables).

Thank you
Adamu

[Carlo Lazzaro]

Adamu:
- -logit- seldom -outperforms -xtlogit- when it comes to panel data analysis (please, see my previous reply);
- see, for instance, -help margins- for postestimation;
- as per FAQ, please note that your chances of getting helpful replies are conditional on posting what you typed and what Stata gave you back (within CODE delimiters, please). Thanks.

[Adamu Idris]

Hello Carlo
Thank you so much for your suggestions.

[Adamu Idris]

Dear Carlo,
sorry to bring back the earlier discussions with regards to logit and xtlogit. Pls can you assist me with any reference material(s) on your argumen that 'logit- seldom -outperforms -xtlogit- when it comes to panel data analysis' as you mentioned?
Thank you
Adamu

[Carlo Lazzaro]

Adamu:
sure: please see Example 1, -xtlogit- entry, Stata .pdf manual.

[Adamu Idris]

Thank you once agan Carlo.