Thread: Analyse quiz data

Hello Everyone,

I am trying to analyse quiz data to work out the winning department.

-------------------------------------
Assume there are 5 departments:

finance 10 employees
HR 20 employees
engineering 100 employees
maintenance 10 employees
operations 50 employees
-------------------------------------

every employee is issued a quiz which they fill in. In the quiz they get plus points for correct answers and negative marks for wrong answers.

what is the best way to find the winning department?

• multiply the results for each dept by a certain factor to make each department contain 100 employees?
• calculate the mean for each dept and select the highest?
• some other better way?

TIA

Originally Posted by chumbawumba

what is the best way to find the winning department?

• multiply the results for each dept by a certain factor to make each department contain 100 employees?
• calculate the mean for each dept and select the highest?
• some other better way?

TIA

I don't think there is any "best" way, but there's probably a way that will incite the least amount of confusion on behalf of your colleagues .

As far as I know, the first two options would give you the same result, but the second will be easier to explain ("the average score for HR was...", etc.).

You could also do things like find out who got the most perfect scores per capita, the fewest abysmal scores per capita, the highest median, etc. None of them could be described as "best", because each might suit a different purpose.

If I were you, I would go with calculating means.

work out which department you want to win, then choose the method that gets you the result you want.

Lies, damned lies and statistics.

Haha, as always Bruxism has the best answer.

Originally Posted by chumbawumba

I don't think there is any "best" way, but there's probably a way that will incite the least amount of confusion on behalf of your colleagues .

As far as I know, the first two options would give you the same result, but the second will be easier to explain ("the average score for HR was...", etc.).

You could also do things like find out who got the most perfect scores per capita, the fewest abysmal scores per capita, the highest median, etc. None of them could be described as "best", because each might suit a different purpose.

If I were you, I would go with calculating means.

thanks for the advice. indeed the first option produces the same answer as the second

I will go for the mean option then, but will standard deviation help in the event that two departments score almost identically?

Originally Posted by bruxism

work out which department you want to win, then choose the method that gets you the result you want.

Lies, damned lies and statistics.

lol

Originally Posted by chumbawumba

thanks for the advice. indeed the first option produces the same answer as the second

I will go for the mean option then, but will standard deviation help in the event that two departments score almost identically?

Probably not. Suppose you have two identical means, but one group has scores with a high standard deviation and the other has a low standard deviation. It must be that one group has some real idiots who are balanced out by some real geniuses, while the other group has no superstars but also no laggards. Which group is better?

I suppose no matter what you decide in the above scenario -- the first group is best, the second group is best, or both groups are equal -- you'll be making a value judgment that has nothing to do with the statistics. So it's more about what you're looking for than what is "right"... hence Bruxism's quote.

Here's where standard deviation stats might really lead you to the right answer: you are the head of an organization whose superstar employees earn for the company many multiples of what the laggards lose. In that case, you might prefer to have a high variance team rather than a low variance team. Does this apply to your company?

thanks for the information.