Negative Chi Square Bartlett;s test of Sphericity

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Negative Chi Square Bartlett;s test of Sphericity

msherman
Dear LIst: I just ran an EFA with 48 variables using SPSS 26 and obtained the following statistics
Kaiser-Meyer-Olkin Measure of Sampling Adequacy = .701
Bartlett's Test of Sphericity   chi square =  -52.441, df =  1035,  sign=  1.00
So the KMO suggests that a factor analysis is appropriate but the correlation matrix is an identity matrix. 
The test of sphericity suggests that I have an Identity correlational matrix and thus factor analysis is not warranted. 
But why is the chi squared negative?
I have tried to search for when one would get a negative chi square for the Bartlett testy but have
been unsuccessful so far. Can anyone point me in the direction to find out why the Chi squared value is negative. 
thanks,  martin sherman
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Reply | Threaded
Open this post in threaded view
|

Re: Negative Chi Square Bartlett;s test of Sphericity

Jon Peck
It looks to me, based on the df, that there are actually 46 variables (df = p*(p-1)/2 but leaving that aside, the Bartlett statistic is only approximately chi squared, and it is computed as
-log(detR) *(n -1 - (2*p + 5)/6)
where detR is the determinant of the correlation matrix, n is the number of cases, and p is the number of variables.

So, with such a large number of variables, the expression can certainly be negative, which is consistent with the p value of 1.

On Wed, Dec 18, 2019 at 12:12 PM Martin Sherman <[hidden email]> wrote:
Dear LIst: I just ran an EFA with 48 variables using SPSS 26 and obtained the following statistics
Kaiser-Meyer-Olkin Measure of Sampling Adequacy = .701
Bartlett's Test of Sphericity   chi square =  -52.441, df =  1035,  sign=  1.00
So the KMO suggests that a factor analysis is appropriate but the correlation matrix is an identity matrix. 
The test of sphericity suggests that I have an Identity correlational matrix and thus factor analysis is not warranted. 
But why is the chi squared negative?
I have tried to search for when one would get a negative chi square for the Bartlett testy but have
been unsuccessful so far. Can anyone point me in the direction to find out why the Chi squared value is negative. 
thanks,  martin sherman
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD


--
Jon K Peck
[hidden email]

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD