Variance comparison between subgroups

[Victor Larsson]

I have reviews from several (domestic) sources for a large sample of films in my dataset. Further, I have specified if the film is foreign or domestic.

So, for each film, I have a review from source 1, one from source 2, one from source 3 (and so on) and a dummy variable, taking the value 0 if the film is domestic, and 1 if it is foreign.

What I want to do is to see if the variance in the reviews of domestic films are larger than foreign films to test the hypothesis that domestic film critics are affected by foreign film critics.

Is there a simple way of doing this?

[Nick Cox]

How does differing variance show an effect? I suspect that husbands' heights vary more than wives' heights, but that is not (e.g.) my wife's height affects mine. I was this height when we met (modulo a slight age effect). Choose your own example.

One way to approach a test of several variances would be to work with deviations on some scale that is zero or positive and to push those through an ANOVA or generalised linear model. You'd need to choose the scale carefully if it was an ANOVA. Something like this seems to be behind various nonparametric tests.

[Victor Larsson]

Maybe i phrased it in a confusing way. All I want to see is if there is a larger spread (on average) in the reviews of domestic films, compared to foreign films. I do not need to show an effect.

[Rich Goldstein]

help sdtest

[Victor Larsson]

If I understands things correctly, sdtest only lets me combine the Std. Dev. for two variables for the whole sample or to compare the Std. Dev. for one variable for each subgroup.

Is there a way of fitting more variables in one test? For example: sdtest source1 == source2 == source3...

[Rich Goldstein]

your #5 strikes me as very different from your #3 - do you want compare the variability in reviews of foreign films to the variability in the reviews of domestic films (#3) or do you want something else? if you want something else, please clarify because I do not follow your #5 - please read the FAQ and post some example data (using -dataex- and within CODE blocks as described in the FAQ)

[Victor Larsson]

I'm sorry. I want to compare the variability in reviews of foreign films to the variability in the reviews of domestic films.

[Rich Goldstein]

sounds as though the issue is in your data setup - however, since you don't show us what that is, I can't help (sounds as though it is in wide format and needs to be -reshape-d to long format, but that's just a guess); I repeat, show some of your data using -dataex-

[Victor Larsson]

This is an example of my data:

416 3 3 2 0
293 3 3 2 0
296 4 3 4 0
664 3 3 3 0
469 3 4 . 0

Where the first column is the number of the film, the three after that is review scores from three different sources and the last is the category variable (in this case only domestic films).

Sorry if I'm being vague in my posts. This is the first time I'm doing anything similar, and I'm kind of confused.