A Regression with population data set

[Nedzad Polic]

Dear Statalist participants,
As I read the textbooks about regression, authors usually assume sample observations availability. Not much has been said, in the textbooks I have read, about the regression on entire population scope of data. Since I am currently working on the population data set of companies in Bosnia and Herzegovina (selected figures from financial reports from 2008 till 2016), namely panel data set with all companies included, if I understand properly my task is to:
1) correctly specify the model,
2) select proper functional form,
3) check for critical assumptions of the model (CLRM assumptions if linear model used),
4) test which approach is more appropriate, either FE or RE model and
5) interpret the results of the regression, since all coefficients are true population coefficients.
Am I right or there is something else I have to take care when dealing with the population data set? Can somebody contribute? Many thanks!

Nedzad

[Carlo Lazzaro]

Nedzad:
welcome to this forum.
You can pretend that your population is sampled from a larger superpopulation (see https://www.cengage.com/c/sampling-d...lysis-2e-lohr: 41-42).
As far as the other questions are concerned, it is impossible (for me, at least) to reply in detail; I would refer you to Stata .pdf manaula; any decent panel data econometrics textbook (see, for Stata users, https://www.stata.com/bookstore/micr...metrics-stata/) and, last but not least, the literature in your researc field.

[Nedzad Polic]

Dear Carlo,

Thank you for you reply and suggestion as well as for a warm welcome.
The list of other questions I posed was related to the premise of the true population coefficient. But, according to your suggestion I should perform all tests related to the coefficient estimates as well. Anyway, very useful comments and suggestions. Thank you once again.