

Thresholds and complex panel data 

Message/Author 


I am using panel data that used a complex (i.e, stratified) sampling design. I want to calculate simple %s (say % male) and get confidence intervals (CI) for the estimated %s. While using the appropriate syntax (i.e, type=complex, a stratification variable, assigning gender as a nominal variable, and including CINTERVALS in the output command) can I just estimate the mean of gender, which is actually a threshold, and use the CI of the threshold to determine the CI of the percentage? Of course the CI's and the Threshold itself would have to be converted using a z table. But would this be an accurate way of calculating the variances/CI's of binary categorical variables collected via a complex sampling design? Thanks. 


If you put gender on the CATEGORICAL list and use the default weighted least squares estimator, you will obtain a confidence interval for the threshold which you can convert using a ztable. 

Back to top 

