Hi all,
I am pulling my hair out over here trying to figure out how to use aweights, I have only ever used pweights until now. I have a dataset with 1,502 participants. The survey platform (Dynata) obtained a sample from a panel of registered voters to gather data using numerous recruitment methods (not relevant here). They oversampled parents, our primary demographic variable of interest. The final sample has 751 parents and 751 nonparents. When accounting for the oversample of parents using aweights, the percentage of parents (n=391, 26%) goes down and nonparents (n=1,111, 74%) goes up to account for the population of registered voters. I have scoured the internet, statalist, and journal articles that have used aweights, and I am still left without a clear direction on where to go. I have contacted colleagues that I know and we are all stumped as none of us have enough training with this kind of weight in analyses. While I understand what analytic weights are and why they are useful in this context (i.e., oversampling), I am having trouble moving beyond simple descriptive statistics and I want to run bivariate and multivariate analyses.
Bivariate: The analyses work for me, but I am unsure of how to ascertain significance. -chi- and -wald- do not work as options. Do you know how I can test significance on bivariate analyses using aweights?
Multivariate: I see that the -logit- command does not work using aweights, and I have read that I must use the regress command. From my understanding, this is because data with aweights are averages. But, I must say I do not really know exactly what that means in general and in terms of interpreting the output. Any help here would be great. I'd gladly read a book, article, etc.
Here is an example of using the command regress to predict a binary DV using two binary IV's (note: it does not contain all my IVs):
First, is this the right command to use? Second, am I missing an "option" command that I could use? If this is correct, could you help me understand how to interpret this?
Thank you so very much! I patiently await your reply,
Brandie
I am pulling my hair out over here trying to figure out how to use aweights, I have only ever used pweights until now. I have a dataset with 1,502 participants. The survey platform (Dynata) obtained a sample from a panel of registered voters to gather data using numerous recruitment methods (not relevant here). They oversampled parents, our primary demographic variable of interest. The final sample has 751 parents and 751 nonparents. When accounting for the oversample of parents using aweights, the percentage of parents (n=391, 26%) goes down and nonparents (n=1,111, 74%) goes up to account for the population of registered voters. I have scoured the internet, statalist, and journal articles that have used aweights, and I am still left without a clear direction on where to go. I have contacted colleagues that I know and we are all stumped as none of us have enough training with this kind of weight in analyses. While I understand what analytic weights are and why they are useful in this context (i.e., oversampling), I am having trouble moving beyond simple descriptive statistics and I want to run bivariate and multivariate analyses.
Bivariate: The analyses work for me, but I am unsure of how to ascertain significance. -chi- and -wald- do not work as options. Do you know how I can test significance on bivariate analyses using aweights?
Multivariate: I see that the -logit- command does not work using aweights, and I have read that I must use the regress command. From my understanding, this is because data with aweights are averages. But, I must say I do not really know exactly what that means in general and in terms of interpreting the output. Any help here would be great. I'd gladly read a book, article, etc.
Here is an example of using the command regress to predict a binary DV using two binary IV's (note: it does not contain all my IVs):
First, is this the right command to use? Second, am I missing an "option" command that I could use? If this is correct, could you help me understand how to interpret this?
Thank you so very much! I patiently await your reply,
Brandie
