I am doing a mulivariable regression to assess which variables are associated with my outcome. Data were collected from 3 schools thus I have 3 PSUs. My question is: is it necessary to control for clustering within these? Since there are so few PSUs.
I have controlled for clustering using the svyset command but it gives me large confidence intervals and changes the interpretation of the results substantially compared with not using this and doing a regular regression.
Any help and advise on this would be hugely appreciated!
I have controlled for clustering using the svyset command but it gives me large confidence intervals and changes the interpretation of the results substantially compared with not using this and doing a regular regression.
Any help and advise on this would be hugely appreciated!