Dear Stata users,
I am trying to perform a multinominal logistic regression using survey data and having trouble with the survey settings.
Would anyone be so kind and help me with the command structure to get svyset right? I consulted the material available and tried to apply it but given my limited experience with the procedure I am worried I might have misunderstood something undoing my hard work (yep, I am a worrier).
About the data: The (crossectional) data is supposed to represent noninstitucionalized national population. To do that households were sampled using stratified two step random sampling which according to the documentation is supposed to be comparable to cluster sampling. There are variables stata (reffering to step 1 when sampling the households) and psu (reffereing to sampling step 2) in the dataset, to correct standard errors making it as though the data were sampled randomly. Within those households (identified by h_id) members were surveyed (identified by i_id). My unit of analysis are those individuals. There are no design weights, only post stratification weights both on households (H_weight) and individuals (I-weight).
As far as I am concerned, there are too many things to go wrong. Could anyone, please, help me out with the command structure for svyset (for this particular case)? I tried to apply notes given on ther situations, but with little luck.
Thank you very much, Anna.
I am trying to perform a multinominal logistic regression using survey data and having trouble with the survey settings.
Would anyone be so kind and help me with the command structure to get svyset right? I consulted the material available and tried to apply it but given my limited experience with the procedure I am worried I might have misunderstood something undoing my hard work (yep, I am a worrier).
About the data: The (crossectional) data is supposed to represent noninstitucionalized national population. To do that households were sampled using stratified two step random sampling which according to the documentation is supposed to be comparable to cluster sampling. There are variables stata (reffering to step 1 when sampling the households) and psu (reffereing to sampling step 2) in the dataset, to correct standard errors making it as though the data were sampled randomly. Within those households (identified by h_id) members were surveyed (identified by i_id). My unit of analysis are those individuals. There are no design weights, only post stratification weights both on households (H_weight) and individuals (I-weight).
As far as I am concerned, there are too many things to go wrong. Could anyone, please, help me out with the command structure for svyset (for this particular case)? I tried to apply notes given on ther situations, but with little luck.
Thank you very much, Anna.
Comment