One strategy would be take the data into EpiData Analysis and find the possible problems there - not in entry. assume: village: vi household : hs (actual number of the household in the village, eg a number from 1-1200) member: mb (1:female, 2: male): mosquito net: mqn
in Analysis: read villagefile sort vi hs mb
*Strategy 1: * generate variable with control value gen i ok = 1 labelvalue ok /1=yes /0=No
* we can look at the previous record with the [_n-1] specifier. if (mqn <> mqn[_n-1]) and (vi = vi[_n-1]) and (mb = 2) and (mb[_n-1]=1) then ok = 0 * now list those records with errors: list if (ok <> 0) or (ok[_n+1] = 0)
* strategy 2: use stattables: * table per household: Stab mqn /by="vi, hs" /stat="min,max" /close * now the "file" in memory is replaced by a new file, where the unit is village household: * list discordant pairs: list if maxmqn <> minmqn
Once you have identified the records with problems, you can find the forms and change the data accordingly in EpiData Entry
regards Jens Lauritsen
On 07/09/2010 15:36, epidata-list@lists.umanitoba.ca wrote:
Hello, I have a questionnaire that i got from interviewing household members. However i was making sure that from each household i interview a male and a female respondent. You realize that some of the questions here for a household have got have the same answers eg Number of persons in the household, number of mosquito nets in the household etc. So, am inquiring to know whether there is any command i can include in my chk file to identify error between two records. I badly need your help, otherwise its so tiresome to check over 1500 records pair by pair. Thank you and regards
Edson B Statistician River Blindness Elimination Program, Uganda _______________________________________________ EpiData-list mailing list EpiData-list@lists.umanitoba.ca http://lists.umanitoba.ca/mailman/listinfo/epidata-list