I have the Kenya Demographic Household Survey and Afrobarometer Survey. Both are national surveys on Kenya on different sets of issues of interest.
There are variables within the KDHS that I am interested and but are not in the Afrobarometer Survey and there are also some variables in the Afrobarometer Survey that are not in KDHS yet I am interested in them. My study needs these two sets of variables in one dataset for analysis.
First, is it possible to combine these two datasets? Secondly, how can this be done in Stata or SPSS (I would prefer Stata methodology for this though)?
Dear Fredrick,
If you combine these data sets together, the results you may obtain wouldn't be trust-worthy. Even if you find significant results, the reliability of the test is under question.
Therefore, I strongly don't recommend to combine these data sets together. You may analyze them separately and combine the information only.
Good luck
Thanks, @Mani Mehraei. Makes sense that way. Is that the standard practice - to analyse data separately then combine the results? I have always seen papers where authors analysed two different sets of data but I have always imagined that they combined the two of them during analysis.
Dear Fredrick,
You only combine or merge datasets if they are studies on the same people and therefore have common variable identifier(s). Otherwise, the data you are bringing together do not cross each other. If they do not have a factor that is common between the two datasets, there is no overlap between the two, so merging has no purpose
Regards.
Dear Fredrick,
Statisticians use data for different populations to compare those populations together. They never merge them together as a one population.
You may use the results out of various populations to make your point. However you should always use hedging because the study doesn't prove anything in statistical point of view.
Good luck.
I have worked with data sets. Combining samples is problematic for validity and reliability reasons. You will find out that these tests will yield different results for each sample. It is best to do the analyses for each data set and then interpret the results and put them in context..
The ability to combine datasets depends on what kind of information you have. For example, if you can match people directly (eg, if one person completed both surveys) it would be great, and people do that all the time. People also match units of analysis with summary statistics,, as when cell phone data is tied to geographical data to improve demographic studies; or when twitter data is combined with county-level health data in the US to predict heart disease.
There certainly can be problems, but these can be identified (and sometimes fixed) through careful application of statistics.
Though the datasets focused on Kenya, however their methods of data collection and questionnaires might not be same.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey