PovertyAction / de_audit

Program to automate sampling for data entry auditing.
0 stars 0 forks source link

Add functionality to sample blanks #2

Closed mbomby closed 8 years ago

mbomby commented 8 years ago

Currently the program only samples from entries that have non-missing values. I'd like to add functionality to either: 1. separately sample blanks or 2. add the option to sample from all values.

mbomby commented 8 years ago

An alternative is to separately sample a (specified) list of key filter questions. Then instead of selecting from a large number of blank answers that are skipped because Question X was "No", you just collapse that and check X itself. This works only if the data entry program was programmed with routing constraints.

mbomby commented 8 years ago

I decided to give two options:

  1. (Default) Sample only non-blanks. You must specify which values count as blanks. These ideally are automatically generated by the data entry program when a section is skipped.
  2. Sample all cells. This is done by specifying the option "blanks"

The data entry guide was also updated to suggest sampling only non-skipped questions, and separately auditing key filter questions, which can be done for a sample of surveys using the -sample- command.