What is the "overall" percentage of a subpopulation for a specific trait?

[dougr]

this is a real life situation. It *is* somewhat difficult to state... one of the reasons (likely) why I cannot envision the answer.



An error log captures incidents for a 24 hour period. Today, for example, there was a total of 155 errors. The total number of clients ("total population") is 3175. Of these 3175 clients, 1053 of them belong to a specific "type" of client. The remaining 2122 belong to a second "type" of client. You could think of them as RED and BLUE, respectively. The RED clients produced 62 errors, the BLUE clients produced the remainder (93 errors). Therefore the RED clients produced 40% of the total number of errors from today.


However, since the RED clients only consist of 33% of the total population, their apparent contribution to the "overall" error rate is higher than then BLUE population, proportionally. The question is, quantitatively, how much higher?

 

[Ishuda]

this is a real life situation. It *is* somewhat difficult to state... one of the reasons (likely) why I cannot envision the answer.



An error log captures incidents for a 24 hour period. Today, for example, there was a total of 155 errors. The total number of clients ("total population") is 3175. Of these 3175 clients, 1053 of them belong to a specific "type" of client. The remaining 2122 belong to a second "type" of client. You could think of them as RED and BLUE, respectively. The RED clients produced 62 errors, the BLUE clients produced the remainder (93 errors). Therefore the RED clients produced 40% of the total number of errors from today.


However, since the RED clients only consist of 33% of the total population, their apparent contribution to the "overall" error rate is higher than then BLUE population, proportionally. The question is, quantitatively, how much higher?

I would think that normally you do individual errors to see which is 'better'. That is, the 1053 RED team had an error rate of 62/1053 and the BLUE team had a 93/2122 error rate.

 

[dougr]

I would think that normally you do individual errors to see which is 'better'. That is, the 1053 RED team had an error rate of 62/1053 and the BLUE team had a 93/2122 error rate.

Yes... thanks you. I was incorrectly trying to associate one population with the other. Your answer is exactly right!