Grouped Data Zero Data Intervals

[Trinkle]

Performing Crow-AMSSA Grouped Data analysis. My data set has many intervals with zero failures. Should I include those intervals? When I do include the zero-failure intervals, I get very different results than when I exclude those intervals.

[tarik]

Yes you should!
Let’s say you are counting failures every 20 days for 60days. And that you have these failures data:
3 failures from 0 to the 20th day
0 failures from 20th day to the 40th day
4 failures from the 40th day to the 60th day.

If you didn’t include the 20-to-40 interval in your analysis and only entered the data into RGA in the following way :
3 failures from 0 to the 20th day
4 failures from the 20th day to the 60th day

Then you lost some insights about when those last 4 failures happened (you are saying that 4 failures happened some time between the 20th and 60th day, whereas you know for a fact that those failures happened between the 40th and 60th day). So you are introducing more uncertainties that you can avoid by saying that you had no failures between the 20th and 40th day.