Hi all,
I am working with paid panel data that was heavily weighted towards white, female respondents. I would like to weight the data to reflect census estimates better using ipfraking but am having some trouble.
First, the ipfraking command requires initial pweights. I have simply assumed a random sample and generated a probability weight equal to 1/pop, (pop = the census estimate of the adult population in the state we sampled). Is that the right initial pweight to use, given that I really do not have a random sample here?
Second, currently I am raking on two variables: gender and hispanic. My code is below:
Running this, I get the error code "Warning: the controls mfemale did not match" and also "The worst relative discrepancy of .86978 is observed for female == 1 Target value = 1960401; achieved value = 255290.16"
Can anyone provide some insight into why there is such a large discrepancy between the target and achieved values for gender? Weights ending up with a value of 255290 for female would imply that 6% of the weighted population is female, wouldn't it?
Thanks for any help.
I am working with paid panel data that was heavily weighted towards white, female respondents. I would like to weight the data to reflect census estimates better using ipfraking but am having some trouble.
First, the ipfraking command requires initial pweights. I have simply assumed a random sample and generated a probability weight equal to 1/pop, (pop = the census estimate of the adult population in the state we sampled). Is that the right initial pweight to use, given that I really do not have a random sample here?
Second, currently I am raking on two variables: gender and hispanic. My code is below:
Code:
* Gender matrix mfemale = (1787688, 1960401) matrix colnames mfemale = _c:0 _c:1 matrix coleq mfemale = _one matrix rownames mfemale = female matrix list mfemale * Hispanic matrix mhispanic = (367239, 120850) matrix colnames mhispanic = _c:0 _c:1 matrix coleq mhispanic = _one matrix rownames mhispanic = hispanic matrix list mhispanic gen pw=1/3748089 ipfraking [pweight=pw], generate(rakedwgt) ctotal(mfemale mhispanic) iter(5000)
Running this, I get the error code "Warning: the controls mfemale did not match" and also "The worst relative discrepancy of .86978 is observed for female == 1 Target value = 1960401; achieved value = 255290.16"
Can anyone provide some insight into why there is such a large discrepancy between the target and achieved values for gender? Weights ending up with a value of 255290 for female would imply that 6% of the weighted population is female, wouldn't it?
Thanks for any help.
