Hello,
This is my first post here, and I'm very glad to join Statalist. I've been a passive user of the forum for a year and a half now, and I've always been impressed with how supportive this community is. Grateful to its members and encouraged by their generosity, I am posting my first question. I hope it won't violate the forum's guidelines, but if yes, I'll apologize and I'll be quick to learn.
My main problem:
- My dataset comes from an online survey conducted in Poland. It has an unbalanced sex and age structure with respect to the population it's supposed to represent. I would like to calculate weights for each gender-age group, separately for each region. To that end, I have downloaded Poland's census data. Then, I would like to apply the weights in my regression models (OLS, robust).
A problem of secondary importance:
- Poland's census municipality-level data only contains the number of people by sex-age groups, but not by sex-age-municipality groups (where, by municipality I mean the type of municipality based on its urbanization and size). Given this, is it possible for me to weigh my observations so that they match not only the sex-age structure of the population, but also the municipality-type structure of the population at the same time?
Questions:
1. Is my calculation formula for weights correct given my regression model and the situation described?
2. Am I using the right Stata command, given the calculation formula?
My weight calculation (applied to each subsample of sex-age-region separately):
prob_inverse = 1 / census_share
where:
census_share = (number of people of sex X, age range Y-Z, region A) / (population of region A)
My stata command:
(Alternatively, I have considered the following calculation:
weight = census share / sample share
where:
census_share is the same as above;
sample_share = (respondents of people of sex X, age range Y-Z, region A) / (all respondents from region A)
However, I am not sure which Stata command the latter weight would correspond to.)
I would really appreciate any help--not only direct answers, but also reading recommendations (preferably a practical guide that I can read and apply quickly). Thank you.
This is my first post here, and I'm very glad to join Statalist. I've been a passive user of the forum for a year and a half now, and I've always been impressed with how supportive this community is. Grateful to its members and encouraged by their generosity, I am posting my first question. I hope it won't violate the forum's guidelines, but if yes, I'll apologize and I'll be quick to learn.
My main problem:
- My dataset comes from an online survey conducted in Poland. It has an unbalanced sex and age structure with respect to the population it's supposed to represent. I would like to calculate weights for each gender-age group, separately for each region. To that end, I have downloaded Poland's census data. Then, I would like to apply the weights in my regression models (OLS, robust).
A problem of secondary importance:
- Poland's census municipality-level data only contains the number of people by sex-age groups, but not by sex-age-municipality groups (where, by municipality I mean the type of municipality based on its urbanization and size). Given this, is it possible for me to weigh my observations so that they match not only the sex-age structure of the population, but also the municipality-type structure of the population at the same time?
Questions:
1. Is my calculation formula for weights correct given my regression model and the situation described?
2. Am I using the right Stata command, given the calculation formula?
My weight calculation (applied to each subsample of sex-age-region separately):
prob_inverse = 1 / census_share
where:
census_share = (number of people of sex X, age range Y-Z, region A) / (population of region A)
My stata command:
Code:
reg y x [pweight = prob_inverse], r
(Alternatively, I have considered the following calculation:
weight = census share / sample share
where:
census_share is the same as above;
sample_share = (respondents of people of sex X, age range Y-Z, region A) / (all respondents from region A)
However, I am not sure which Stata command the latter weight would correspond to.)
I would really appreciate any help--not only direct answers, but also reading recommendations (preferably a practical guide that I can read and apply quickly). Thank you.
