Linear regression with repeated measures

#1
Hi, I have the following question:
I have a dataset that contains data from 9 subjects that were tested twice for a couple variables. I try to do a linear (multiple) regression on the data, but I understand that it is not ok to simply pool the data.

If I average the two measurements, there is no chance for significant correlations. I tested the repeated measurements for pearson correlation, but found no significant R for any, thus it would almost appear as if the data is more independent than perhaps expected from the way it was obtained.

Is there a way to correct for the repeated measures without using an average i.e. could I change the df or other variables to accommodate for this?
 
P

pad1w07@soton.ac.uk

Guest
#2
Hi, I have the following question:
I have a dataset that contains data from 9 subjects that were tested twice for a couple variables. I try to do a linear (multiple) regression on the data, but I understand that it is not ok to simply pool the data.

If I average the two measurements, there is no chance for significant correlations. I tested the repeated measurements for pearson correlation, but found no significant R for any, thus it would almost appear as if the data is more independent than perhaps expected from the way it was obtained.

Is there a way to correct for the repeated measures without using an average i.e. could I change the df or other variables to accommodate for this?

you can use pooled OLS ( multiple regression) on panel data, but it is usually not optimal, try using fixed effects or random effects
 

FAST5

New Member
#3
Why isn't pooled OLS optimal? If you control for serial correlation and heteroskedasticity (robust command in stata) is pooled OLS still preferred over random effects?