Can we apply different estimation techniques in a single paper?

[tauseef ali]

In a research paper, if we use, for example, the ZINB model for our main analysis (main table). Can we use another model such as ZIP or NBREG for the rest of the robustness tables? I mean, if I use ZINB for table 3, so can I use ZIP or NBREG to generate table 4 (i.e., by using another DV proxy), table 5 (i.e., by using another IV proxy), and table 6 (i.e., by using another sample), etc.?

[Nick Cox]

What's allowed is a matter for any gatekeepers judging your work, whether advisors, supervisors, examiners, reviewers, editors, whoever. .

But there must be thousands and thousands of papers comparing different models and different methods, and the broad goal there is surely good. How well you do it is the deal.

Compare https://www.statalist.org/forums/for...-with-outliers

[tauseef ali]

What's allowed is a matter for any gatekeepers judging your work, whether advisors, supervisors, examiners, reviewers, editors, whoever. .

But there must be thousands and thousands of papers comparing different models and different methods, and the broad goal there is surely good. How well you do it is the deal.

Compare https://www.statalist.org/forums/for...-with-outliers

Thank you so much, professor, for your reply. Actually, the purpose of my research paper is not to compare models. I want to switch models from one to another because the first model is more preferable in literature for some specific kinds of the dataset but hard to have significant results. So, my question is that if I just want to switch the estimation technique for the remaining robustness analysis, is it allowed, or should I continue with one estimation model in the whole paper

[Nick Cox]

It’s the same question really, but let’s answer differently. Whatever papers appear in your literature are evidently allowed. Perhaps you’re planning something original, in which case you will need to convince others that you have something good as well as different.

[Roman Mostazir]

Honestly speaking, I don't think this is a matter that anyone can advise you on unless the person is affliated with what you are doing. Without knowing your research, it is difficult to comprehend why are you trying to use different estimators. Of course there are papers which compares results of different methods/estimands which are perfectly valid but your's seem don't fall into that line. Choice of an estimand comes from data distribution type and hypothesis being explored but certainly not from the fact that 'I like the result of one than the other'. You need to discuss this with your fellow colleagues and establish the firm ground in your paper for using different estimators, otherwise the reviewers may doubt the results as 'cherry picking'.

[Just cross posted with @Nick Cox]

[tauseef ali]

Honestly speaking, I don't think this is a matter that anyone can advise you on unless the person is affliated with what you are doing. Without knowing your research, it is difficult to comprehend why are you trying to use different estimators. Of course there are papers which compares results of different methods/estimands which are perfectly valid but your's seem don't fall into that line. Choice of an estimand comes from data distribution type and hypothesis being explored but certainly not from the fact that 'I like the result of one than the other'. You need to discuss this with your fellow colleagues and establish the firm ground in your paper for using different estimators, otherwise the reviewers may doubt the results as 'cherry picking'.

[Just cross posted with @Nick Cox]

Thank you professor.

[tauseef ali]

It’s the same question really, but let’s answer differently. Whatever papers appear in your literature are evidently allowed. Perhaps you’re planning something original, in which case you will need to convince others that you have something good as well as different.

Thank you so much professor.