Within Family fixed effects in cross section

[Paul Shlapa]

Hello everyone,

I have a data for multiple individuals from different families at one (1) point in time. As i do a replication study, one of the approaches by the authors was to use a family-fixed effects to estimate difference in height using a within family comparison, assuming that nutrition patterns…. were shared by the individuals coming from the same household.
A question is how to do it properly in stata…

My approach:

xtset orighhid
xtreg height exposed age_days age_days2 sex i.bth_month, vce (cluster orighhid) fe

Is this correct?

[Carlo Lazzaro]

Paul,
I do noth think your approach is correct, as you have one point in time only, so I would use the following code:

reg height exposed c.age_days##c.age_days i.family i.sex i.bth_mont, vce(cluster family)

Please also note that both linear and squared terms for -age_days2_ are now coded more effficiently thanks to -fvvarlist- notation.

[Paul Shlapa]

Thank you Carlo,
I also performed an OLS regression with the inclusion of dummies for families, still the authors imply that they had an alternative specification as random effects, which turned my thought to the initial approach. Also thanks for the tip on age coding

[Carlo Lazzaro]

Paul:
please note that only one specification between -fe- and -re- will be correct.