The value is negative due to a negative average covariance a

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

a.ticlo
Posts: 1
Joined: Wed Aug 13, 2014 5:17 pm

The value is negative due to a negative average covariance a

Postby a.ticlo » Wed Aug 13, 2014 5:19 pm

Cheers! I have searched the entire library and internet and i can still not find a way to solve this:

data set(1)
clustprim.ctxRH.MOF clustin.ctxRH.MOF
0.358974 0.333333
0.488889 0.321429
0.448718 0.355556
0.362637 0.377778
0.410256 0.527778

when I run Analyse-->Scale-> Reliability analysis
then check those 2 items( clustprim.ctxRHMOF.; clustin.ctxRH.MOF)
and check Intraclass correlation coeff, two way random, abs agreement;
I get the error mentioned on the title,
http://i57.tinypic.com/258402f.jpg
Image

Why do I get begative Cronbach alpha and negative intraclass correletation coeff's?
How to solve it? As bad correlated as they could be, it shouln't be begative, am I right?

If I do the same on this:
dataset(2)
clustprim.ctxRH.RMF clustin.ctxRH.RMF
0.571429 0.8
0.362637 0.287879
0.435897 0.418182
0.411765 0.408333
0.366667 0.406593

everything goes perfectly fine..

Regards
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: The value is negative due to a negative average covarian

Postby GerineL » Fri Aug 15, 2014 8:38 am

Most likely, you have negatively phrased items in there (either that or an extremely small sample size).
So if you look at items in a scale, it could be that some items are phrased positively (I love X), and some are phrased negatively (I hate X).
If it is persons you are looking at, it could be that they score opposite. So for instance, if you look at how likely it is you will win a game against one other person, the likelihood of person 1 winning increases when the likelihood of person 2 winning decreases and vice versa.
Could you check this?

Who is online

Users browsing this forum: No registered users and 5 guests

cron