Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
COMBINING THE RESLTS OF TWO SAMPLE TYPES?
#1
Hi Id appreciate some sound advice on whether I can legitimately combine the results of two surveys. Both asked the same questions but one was conducted randomly with the population by a research interviewer (1000 responses) and the other was an online self-selection survey (500 responses).  Both surveys sought to determine the demographics of the population and their use and perceptions of a city centre.  Can anyone point me in the right direction here of whether they can be joined and reported on or if it is totally best to keep them separate. Many thanks
Reply
#2
First thing to do is to make a comparison between results from the two samples. What differences are there between them? If demographics are different, then you can try weighting the self-selection survey to the profile of the random sample and again look for differences. If there are few differences and you feel confident that the weighted self-selection sample is as representative as the random sample then I would say you can combine them (this would be similar to a rep+boost type sampling plan), but I would want to check back to the random only sample for any particularly important findings.

In practice, the combined sample will be of most use where you're looking a smaller sub-segments and don't have enough sample for the random-only method. Since the sample size will be smaller for these types of sub-segments you might want to treat results more as indicative than definitive. Again, if it's important see what random alone would show, or look to confirm the finding with additional research.






Reply
#3
(Thu 24 Nov 2016, 12:09)Saul Dobney Wrote: First thing to do is to make a comparison between results from the two samples. What differences are there between them? If demographics are different, then you can try weighting the self-selection survey to the profile of the random sample and again look for differences. If there are few differences and you feel confident that the weighted self-selection sample is as representative as the random sample then I would say you can combine them (this would be similar to a rep+boost type sampling plan), but I would want to check back to the random only sample for any particularly important findings.

In practice, the combined sample will be of most use where you're looking a smaller sub-segments and don't have enough sample for the random-only method. Since the sample size will be smaller for these types of sub-segments you might want to treat results more as indicative than definitive. Again, if it's important see what random alone would show, or look to confirm the finding with additional research.
Thanks so much for your help :-) Much appreciated Saul.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)