Test for error of central tendency

#1
Dear community,

I have two groups of users who rate on 7-point-scores for n data items. Now I want to test whether there is a difference in both groups in their error of central tendency, i.e.: does one group tend to rate more towards the middle of the scale than another?

I assume the ratings for one data item and both groups together to be distributed according to a normal. Can you recommend me the best test for this task? I don't know if an ANOVA is appropriate because it compares means, not variances within and between groups.

Thank you very much in advance,

Christian
 

Karabiner

TS Contributor
#2
Why do you think your question is "excellent" (five stars)? Just curious.
AFAICS the postings initially self-rated with five-stars receive less answers,
on average, than other posts, so I am wondering a bit. Since the tendency
to self-rate with 5 stars is most frequent in the psychology subforum, it might
me a matter of undergraduate pop psychology going terribly wrong, but I don't
know...

As to your question, perhaps Levene's test, or perhaps Moses extreme reactions
test.
 
#3
Dear Karabiner,

in fact, I'm not dealing with ratings from "bad" to "excellent", but with property estimations like: "Here is a price ranging from 0..1000. Which price do you estimate for this item"? Just an example. Thus, very high and very low values will certainly be more uncertain than the middle range depending on the item. In my particular experiment, there are two groups of participants and my hypothesis is that one group tends to select less extreme values or, expressed differently, has a stronger error of central tendency. This is what I want to test.

And btw: I'm not a psychologist, so any non-statistical feedback is highly appreciated as well.

Thanks for your comments,
Christian
 

Miner

TS Contributor
#4
I don't know if an ANOVA is appropriate because it compares means, not variances within and between groups.
This is incorrect. ANOVA is a test for effects (utilizing the ratio of between groups variation to within groups variation), not means. That is why Post-Hoc means tests are required to isolate which levels are responsible for the effect.