Hi, I am running a logit regression in Stata. I do not see standard errors and p values in some regressors (for a few models) and in all regressors ( in some other regression models).
Pseudo R square is 1 in the models where SE and p value corresponding to no regressor is reported by the Stata.
I understand that this is happening due to either the problem of separation (quasi or full) or reduced degrees of freedom due to inclusion of higher number of regressors (sort of k>n).
Details for dataset are as follow:
--> Model includes a regressor in the level form as well as the square term (quadratic function).
--> In addition, the model includes 9 control variables plus industry and year controls.
-> Industry includes 34 unique industries in form of 2 digit industry codes.
--> Year control includes 11 unique years. Cross-sectional data is pooled across those 11 years.
If I remove industry effect controls the results become inconsistent in the models. Somewhere these give opposite signs and in some models these become insignificant.
I would be grateful your kind suggestions.
Thanks you!
Try decreasing the number of predictors and see if that helps . What is the total number of observations? Best wishes David Booth
Thank you David Eugene Booth and Hichem Guizani !
Hichem Guizani , Inclusion of more variables (or another form i.e. interaction term) will complicate the model in two ways. First, It will make the interpretation difficult. Second, calculation will be complicated and measurement becomes even more difficult. If I could understand your point clearly.
David Eugene Booth , I have 169 firms with cross-sectional observations.
I have tried reducing the number of predictors by removing industry effects (industry classification codes), I works fine now.
--> But including industry effects seems to be a fashion in the papers. I worry if reviewers put a question mark on this point.
--> secondly, results in one of my models, where convergence is well achieved, main predictor loose the significance if I remove industry effects.
--> I have heard of some other ways of calculating likelihood or gradient descent which I'm looking for hoping that will help.
Best Regards,
Hello Sahil, based on what you just said the problem seems to be too many IVs for your number of observations . I understand that you would like to include industry effects but you don't have enough observations for that big a model. Therefore increased n or decreased number of predictors seems to be your options. Best wishes David Booth
Hi Sahil Narang, From what you described above I assume that the N for your outcome is 169, right? If so, how many observations are in each category of your binary outcome? And how many predictors did you include in the model?
Once I have these answers, I can comeback to you the soonest.
Best wishes
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India