I have a large amount of data that was collected by "convenience sampling," (i.e. it is not "nationally representative."). As such, there may be sampling bias in the statistics calculated from these data. Could I subsample it so that the sample distribution matches, say, the population distribution (as determined by a census, for example?)
For example, say I already measured the body weight of millions of people in the US, and I know the demographics of every person in my sample. I would like to resample my data so that the joint distribution of these covariates matches the joint distribution of the population, which is determined by census data. If my sample has a preponderance of old, rich, white people, then this process will effectively remove enough of these old, rich, white people so that the distribution of age, income, and race matches the census distribution of those variables.
Is this common practice? Does it make sense? If so, how can I go about doing this?