The National Institute of Population Studies (NIPS) in Islamabad, Pakistan, was given funds by the UK Overseas Development Agency (ODA, now the Department for International Development) to carry out a nationally representative survey of fertility and family planning uptake. The work was co-ordinated by the Centre for Population Studies (CPS) at the London School of Hygiene and Tropical Medicine. Andrew Westlake acted as advisor and consultant for the data processing and report production for the survey.
The survey covers some 8,000 households in 250 clusters across the major provinces of Pakistan. Fieldwork was completed at the end of March 1997, with a few less than 8,000 eligible women (ever-married, aged 15 to 49) being interviewed.
Data entry was done with the package PC-Edit produced by the United Nations. The program is not widely used, and does have some problems, but was already known to the computing staff at NIPS. It has one major strength, its ability to handle complex record structures. With it we are able to input the whole of the survey into a single file, including information for Household, Household Member, Eligible Woman and Birth History. The setup of the entry program is straight forward once decisions about the underlying record structures have been made. Repeating records are easily handled in the program code. The programming system is powerful, if somewhat idiosyncratic, but permits complex checks to be written and some field-level imputation. The same program is used for data entry, for verification (of selected fields), for updating, for batch checking and for making error listings. It runs adequately under MS-Dos on 286 PCs, which are the machines most readily available at NIPS.
To supplement the facilities provided by PC-Edit a number of programs have been written in Visual Basic, linked to MS Access under Windows 95. These process listings produced by PC-Edit to extract relevant information so that it can be presented in a more usable form. In this way a data dictionary listing is produced which shows all the fields with their labels, codes, ranges, missing values, not applicable codes and skip patterns. This is being used for checking after data entry, and will be a resource for subsequent users of the data. A similar listing shows all the program code associated with each field, for use by the programmers and as a record when data entry is completed.
The programmers do not have much control over the format of the error listings produced by PC-Edit (only over their content) so we have produced a program to reformat the listings and relate the error messages more usefully to the data records. This is the main material with which the data editors work.
All the elements of the processing system were in place and operational soon after the start of fieldwork, but progress was slow because of difficulties in finding sufficient manpower for the various operations. The main bottleneck is with the programming staff (only two) who are needed to perform the updating operations on the data file after checking and cleaning. Their time is in great demand from other projects within NIPS. There was also a problem in finding enough skilled people for the checking to keep up with the data entry.
PC-Edit has a companion program called X-Table which produces simple tabulations from the main data file. This was used to produce intermediate statistical reports for the survey managers. Final tabulations are produced using SPSS 8 under Windows 95, to which PC-Edit will export its data (together with the data dictionary information). From SPSS we can transfer the results of analysis into other Windows programs.
The majority of the preliminary report from the survey was written in advance, since the items to be reported flow directly from the questions which motivated the content of the questionnaire. The comparative figures from other surveys are already there: all that remained was to insert the figures from this survey when the data cleaning was completed, and to review the comments about these results. This report was published in February 1998.
Work on the main report was completed early in 1999, and the published version is available from NIPS in Islamabad. At the same time a CD was produced, to support further analysis. The CD contains the data from the survey (both the cleaned raw data and a reorganised version in SPSS, including useful derived variables), full documentation of the data files (including documentation of the processing stages used to enter, clean and reorganise the data), and copies of the original questionnaire and the final report as Word files. Any requests for access to the data, or for permission to use the data for further analysis, should be directed to CPS in London.
Page last updated 18 Jun 2003.