DIAL RFP: Up to $25,000 USD available to improve ODK

Community
2

Thank you, @yanokwa.

Yes, I love to start extending our earlier work on curbing data entry errors. We could maybe identify one or two possibly problematic question types.

In the earlier experiment we had 24 interviewers enter more than 12,000 known dates under 12 different combinations of factors (phone vs tablet, single- vs double-entry, and three different date widget/interfaces. We found errors in about 10 in every 100 dates using single-entry with default calendar interface on a phone and we found errors in about 2 out of every 100 radio button entries where the choice was simply
o Male
o Female

You have made some subsequent changes to the date widgets. We could design a short instrument and run another experiment learning from last time. Count me as throwing my hat in the ring if this idea bubbles up as one of the top suggestions here.

A second idea would be to focus on several protocols for double-entry and see how the rates compare to single-entry:

  1. One operator enterst the field and enters it again immediately; they must match; this is what we tested before
  2. One operator enters the data in the double-entry items, and then hands the device to their partner who enters them again. For each item in the list, if there is a mis-match, the device asks the second operator to enter the corrected value, and then moves on.
  3. Of course that second protocol could be done by one person. Enter a set of items once, Then the questionnaire has you start again and enter them a second time, and when it finds a mis-match it stops and asks you to enter the definitive value.

We wouldn't want to do that for the entire questionnaire, but maybe for a small set of items that are crucial. In my case it would be for the dates that are written on a child's vaccination card.

We can't answer all the questions for $25k, but we could either run another decent-sized experiment, maybe with two international partners, or we could drill down on double-entry and see if there is a touchscreen equivalent of the keyboard practice that yields similarly low rates.

-Dale

1 Like