Hi! Hope everyone is doing well,
I am currently working with a pseudo panel whose observations come from three cross sections (2014, 2017, 2021) of the LSMS survey from Bogotá, Colombia. I have 91 cohorts that were formed from having 13 birth year and 7 region groups. Some of my cohorts have up to 5000 observations, while others have very few, like 20 or 30. That is explained because most of the people surveyed come from Bogota, and one of my regions (variable I used to create cohorts) is Bogota.
I want to run both, a pooled regression and a fixed effects regression, which will later serve as the inputs for some inequality indexes. Due to that I have few periods, T=3, t C=91 and I don't have a constant number of observations per cohorts, I understand that I must use the approach suggested by Verbeek and Nijman(1993) and not the one suggested by Deaton(1985).
My questions:
i) Is there any package, command in stata to run a regression as the one proposed by Verbeek and Nijman(1993)
ii) Should these regressions be run using weights? what would be the meaning of the weights when using cohorts?
iii) When are the number of cohorts (C) and number of observations per cohort (nc) considered big, small, fixed?
I appreciate any help,
Literature:
Verbeek, M. Nijman, T (1993) Minimum MSE estimation of a regression model with fixed effects from a series of cross-sections.
Deaton, A. (1985), Panel Data from Time Series of Cross Sections, Journal of Econometrics,
I am currently working with a pseudo panel whose observations come from three cross sections (2014, 2017, 2021) of the LSMS survey from Bogotá, Colombia. I have 91 cohorts that were formed from having 13 birth year and 7 region groups. Some of my cohorts have up to 5000 observations, while others have very few, like 20 or 30. That is explained because most of the people surveyed come from Bogota, and one of my regions (variable I used to create cohorts) is Bogota.
I want to run both, a pooled regression and a fixed effects regression, which will later serve as the inputs for some inequality indexes. Due to that I have few periods, T=3, t C=91 and I don't have a constant number of observations per cohorts, I understand that I must use the approach suggested by Verbeek and Nijman(1993) and not the one suggested by Deaton(1985).
My questions:
i) Is there any package, command in stata to run a regression as the one proposed by Verbeek and Nijman(1993)
ii) Should these regressions be run using weights? what would be the meaning of the weights when using cohorts?
iii) When are the number of cohorts (C) and number of observations per cohort (nc) considered big, small, fixed?
I appreciate any help,
Literature:
Verbeek, M. Nijman, T (1993) Minimum MSE estimation of a regression model with fixed effects from a series of cross-sections.
Deaton, A. (1985), Panel Data from Time Series of Cross Sections, Journal of Econometrics,
