Understanding Society User Support

Hi Abdulganiyu Ajani,
you say you are merging files ... and then that you are appending variables ... this makes it difficult to guess what is happening.

"append" and "merge" are different ways of combining information from different waves. E.g. if you merge indresp files for wave 1-18 on pid and keep only those observations that have _merge==3 you'd expect a significant reduction in sample size because only persons who gave a full or proxy interview in all 18 waves would be present in the final merged data set (this number should be less than 4,000 if I remember correctly).

Or are you referring to the number of observations in your models? Note that not all information is available in all waves, and within waves people may not provide the information. Estimation models will only consider cases that have valid entries on all variables included in the model. If a variable is present only once, only information from that wave is used.

You may find it useful to consult the materials in our online training course "Introduction to the BHPS using Stata"; the first three or four exercises in include detailed descriptions of how to merge/ append BHPS files. Please visit
https://www.understandingsociety.ac.uk/documentation/training/online for further information about online training courses.

Hope this helps,
Gundi