Skip to main content

Data analysis

(go to Outline)

Was the sampling done correctly? If the sampling is described in enough detail, the reader can judge whether or not there are potential biases injected by the incorrect sampling. Certainly, the reader should be able to judge if the sampling was done randomly or by convenience.

Were the data analysis techniques appropriate to the sampling method? If a cluster sample was done, did the calculation of confidence intervals take this into account? If not, the confidence intervals calculated are narrower than they should be, leading readers to believe that the survey results are more precise than they really are. This is a very frequent mistake that demonstrates that the survey managers did not know as much about survey statistics as they should have known.

Did all the survey subjects have the same likelihood of selection; if not, was different likelihood accounted for in the analysis? If the same sample size was selected in one refugee camp with 1,000 households and in another refugee camp with 5,000 households, the likelihood of selection is very different. Analyzing these data together without appropriate analytic techniques can potentially introduce sampling bias.