Monday, May 11, 2009

Juking the stats

Here is another true story from work that seems like it was taken directly from an episode of The Office.

Towards the end of each year each school has a survey online that is suppossed to be taken by as many staff, students, and parents as possible. They ask questions like do they enjoy coming to work, do they fell safe, do they feel supported and appreciated, etc. You take or you don't, but they would like as many as possible to take it. I took it in about 8 minutes.

Today at this meeting, it was my job to just once again mention to staff to take it please. I mention this and immediately a staff member has a comment. She is upset that after you answer all the questions, it takes you to another page asking questions about gender, race, and how long you have been at this school. It says that this page does not affect the results, its just for calculation purposes or whatever. She argues that when she got there she canceled out of the survey because she thought who ever looks at these results could easily identify her. She said that she would mark that she teaches English, is white, female, and has been here 8 years; and she is the only teacher that fits this profile. She didn't want to complete this survey because someone could identify her.

At this point I tell her to just give the school and administration great marks on the survey that way if she was identified ( which she never would be), they would see that she thinks great about the school and administration. She then said that if she did that she would invalidate the whole survey, and all the results would be useless.

I then suggested to lie about her demographics on the last page, mark that you are male or have been here for 4 years instead of 8. She then said that this would also invalidate the whole survey for everyone. At this point I gave up. So her decision was to just not take it. Wouldn't this invalidate the whole thing just as much? Would it really be the end of the world if someone identified her survey. Again this would never happen because a computer is calculating these results and then posting them in a pie chart.


Here is a funny skit from SNL on Saturday

2 comments:

edgrimly said...

Close to if not better than "Dick in a Box." We get surveys like that once a month. One month it is all 10 out of 10, the next it is the worst possible. Have to keep them guessing.

Muntaba Lambego said...

http://www.theonion.com/content/node/33736