I'm currently looking at survey results from the US bureau of labor statistics concerning hourly wages. With the raw survey data, I get different respondents who state how many hours they work in a week, along with the average pay, and some demographic data like age and sex.

How can I go from this information to making an educated guess about the amount of hours have been worked by the entire US population at a certain wage level, for example? Is there some way to "correlate" the values considering more than just a single parameter like age?

EDIT: I'm going to try and ask a more specific question then. Please forgive my complete lack of statistics knowledge:

Let's say I'm doing a survey about how much people earn in an hour. Based off of my survey, I know that , in my survey sample, men make about \$10 an hour, and women make \$8 an hour. I also know that in the same sample, people under 40 make \$7 an hour on average, and people over 40 make \$11 an hour.

Assuming that the population is 50/50 male/female, I can make an educated guess that the population , on average, makes \$9 an hour. Likewise, if %30 of the population is over 40, then I can guess that the population on average makes \$8.20 an hours

But how can I "combine" both of these information points? How can I consider both the sex and the age of the people to fit it to the population? Am I just supposed to do a weighted average or something?

  • 1
    $\begingroup$ It is difficult to imagine a useful concise answer to this. See stats.stackexchange.com/help/dont-ask including the advice "Your questions should be reasonably scoped. If you can imagine an entire book that answers your question, you’re asking too much." $\endgroup$ – Nick Cox Jul 7 '13 at 9:10
  • $\begingroup$ ok, I tried to edit my question to be more specific. I fear that it might still be too general. $\endgroup$ – rtpg Jul 7 '13 at 9:41
  • $\begingroup$ just use any survey that generalizes to the whole population instead of the working population.. you probably want the bls current population survey. to work with the microdata, start here $\endgroup$ – Anthony Damico Jul 7 '13 at 13:44
  • $\begingroup$ @Anthony: but, CPS generalizes to the working age population not the whole US population; the survey focuses on the civilian noninstitutional population 16 years and over. $\endgroup$ – Metrics Jul 7 '13 at 14:03
  • $\begingroup$ @metrics that is not correct. this page says, "the civilian noninstitutional population" kids don't get the employment questions, but there are child records - with nonzero weights. it's a household survey :) if you don't believe me, sum up the person-weights -- you'll get ~310 million not ~230 million $\endgroup$ – Anthony Damico Jul 7 '13 at 16:41

To generalize the results from survey data for the population, you need sample weights. If you are using the microdata/survey data from BLS, you can find the weights in the data. Of course, you have to look into codebook or data dictionary if you can't locate. See this introduction slides.

For analysis purpose you need to use survey analysis tool which is this in R and this is Stata


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.