Calculating error rates with independent keying

[BPAnderson]

This question might not belong here in Entry, but rather in Tabulation, but I thought I would start here.

I am trying to find out if there is a way to calculate an error rate for a keyer. I know that the .log file will give the number of bad keystrokes, but this really isn't an error rate since it doesn't track when a keyer enters a valid, but incorrect, response and then goes back and corrects/re-keys it. The log file also has a 'number of fields with errors attributed to keyer/verifyer,' but I believe that is based on dependent keying. I also know that the comparison file will show differences between independent keyers, but it doesn't give error rate information since it doesn't know which independent entry was incorrect.

And so, is there a way to actually determine a keyers error rate based on the number of mistakes made during keying? In this case, mistake would be defined as an incorrect entry that meets the entry criteria (a valid answer), but is the incorrect response based on the data entry form. This would be the type of error found during independent keying.

[Gregory Martin]

Is an independently verified data file available, or are you trying to calculate error rates without a comparison file?

If you are doing independent verification anyway, can you run the Compare Data tool on the original data file and then the verified file, and the number of errors would equal the number of errors the keyer made?

[BPAnderson]

Yes, a comparison file is available and a Compare Data can be run. However, once the number of errors is known from the Comparison file, how is it known how many total entries were made in order to determine the error rate? Counting just the errors in the Comparison file would only say how many errors were made, but not an actual error rate. I have a feeling that a true error rate is not possible when doing independent keying.

[Gregory Martin]

This is not possible in any easy way. As you know, you can get the number of errors from the Compare Data listing report, and you could even generate this automatically by writing a batch program that reads the listing report instead of counting the lines in the report manually, but that does not do you much good unless you have the denominator of the number of fields entered.

CSPro does not store this number, so the only way for you to get this would be to create a batch program to count it yourself, checking whether each field was notappl or not. Needless to say, this would be rather annoying to code, but it is possible.