Industry, country and year effects

[Oreshile Sulaiman Ademola]

Hello everyone!
I need your suggestions and guidance on some Stata issues.
I am currently working on a small sample of 130 African firms from 2013-2018, arranged into panel of (10 industry classifications and 8 countries) and i have created country id (id_country) and industry id (id_industry) respectively in the dataset
I have the following questions;
1. which of the regression estimation model (OLS, FE or RE) should i add these ids (id_country and id_industry)
2. i included 8 control variables and across the three estimation models only 2 are insignificant and varying signs. So what do you suggest i do with these? what is the ideal number of control variables to add since i have a small dataset and also intend to run industry level and country level analysis.
3. According to Hair et al. (2010), there is an ideal number of observation of each regressors. Some authors also suggested (5, 10, 15, 20, and 25) with my dataset having as low as 2 firms from some countries, what are the suggestions you can offer since i intend to run industry and country level analysis

Thank you

[Carlo Lazzaro]

Oreshile:
welcome to this forum.
1) if your want to retrieve both -country- and -industry- -fe-, you should theoretically speakng, consider the community-contributed module -reghdfe-, with the proviso that the -fe- machinery wipes out time-invariant variables;
2) there is neither universal, nor hard and fast rule about that. The idea is to specify the regression so that it gives a fair and true view of the data generating process you're investigating;
3) See https://projecteuclid.org/journals/a...176346793.full

[Oreshile Sulaiman Ademola]

Thank you so much Lazzaro for your swift response.
I went through the link you share but still not clear

[Carlo Lazzaro]

Oreshile:
see https://www.statalist.org/forums/for...ols-regression, #4 and related link.