08 August 2006

Bah.

I've spent the better part of the afternoon trying to remember the proper way to compute a Z-test for proportions. This has not been fun. It doesn't help that the data I'm using is non-representative for gender, so the results need to be considered by gender, which just adds one too many pieces too the puzzle for me to feel like I have any hope of identifying the correct sample size to put into the silly little formula. And while there are calculators out there on the web that will compute this for me, I have a lot of these little buggers to calculate and I don't particularly feel like plugging them all into a little calculator when I *should* be able to just put the damn formula into Excel. *Should* being the operative word there.

Guess I'm making a trip to the Statistical Consulting Center tomorrow. *sigh*

2 comments:

Anonymous said...

Dude, I know I'm starting to sound like a broken record here, but this is one of the only things SPSS Tables is good for. You check a box, and it does diffs (aka z-scores, aka difference of proportions tests, aka column proportions tests) for *your entire table*, comparing every number in your table to every other number in its row. This is how marketing research works--chi square was something I just used in school before I started in IR--and it's very handy because sometimes the thing that you thought was significant from a table isn't the thing that is significant. Anyway. SPSS tables. Dude. There might be a way to find it without the tables module--like I said, they call it a column proportions test.

Anonymous said...

Yeah - I don't use SPSS these days, but SAS would do it in a jiffy too. I'm glad Cate got to you first though and solved the thing in one fell swoop!
-Sara, www.saraskates.typepad.com