TIME SERIES DATA QUALITY CHECKS

[luizaandrade]

The code to do this is very repetitive right now. It seems to me that this could be automated into a function like this:

• Inputs

  • A data set that has already been wrangled (and ideally also labelled)
  • The name of the date variable
  • The name of the enumerator name variable
  • The names of the variable indicating the survey duration
  • A list of variables that you want to check the average response per day
  • A list of variables indicating the number of "other" selected that you want to check the average number of occurrences in a day

• Outputs: graphs with the listed variables (plus average number of surveys and average duration) graphed by day and by enumerator + day

• What the function does

  • Summarizes the data by date and by date + enumerator as in the code below, but taking the variable names as inputs: https://github.com/dime-worldbank/iesurveykit/blob/dda8c5785f2b73e5163a55c2173540e2a7ad219c/Survey%20Checks/r/sample_analysis.R#L125-L148
  • Sets a default graph as in here: https://github.com/dime-worldbank/iesurveykit/blob/dda8c5785f2b73e5163a55c2173540e2a7ad219c/Survey%20Checks/r/sample_analysis.R#L32-L94
  • Graphs all the listed variables
  • Saves the graphs

[luizaandrade]

In addition, I personally find it a lot more useful to see a box plot than averages for duration. In the Moz team, we'd have a box plot of duration over the date and a box plot of duration over enumerators.