Cronbach's alpha

[Larnie]

Hi

I tested a group of 6 parents, who completed a questionnaire, which
contained 15 questions (likert scale response format).

This was part of a pilot study. I used a previously validated measure;
however, the words 'mental health service users' was changed to 'parents
of children with ADHD', to suit my sample. Participants also provided
qualitative data regarding the usefulness of the scale for use with
parents of children with ADHD.

I ran the analysis to determine the Cronbach's alpha for the modified
scale (which will be used in the main study).

Cronbach's was .920

BUT>>>>this warning message comes up on SPSS:

'The determinant of the covariance matrix is zero or approximately zero.
Statistics based on its inverse matrix cannot be computed and they are
displayed as system missing values.'

and no squared multiple correlations were calculated.


What would have caused this to happen?

Parents felt that a few items needed to be modified for various reasons
(hence their responses to these items were not as expected), could the
problem be related to this?

Can I still use the Cronbach's?


Need help, not a stat guru.

Thanks

=====================
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

[Swank, Paul R]

You have 15 questions which generates a variance covariance matrix with 120 unique entries. So you are estimating 120 variances and covariances with 6 subjects. This means a lot of room for chance correlations dependent on the particular sample. I think I would trusty a Guttman-Flanagan split half more in this case.

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Larnie
Sent: Thursday, August 06, 2009 6:07 AM
To: [hidden email]
Subject: Cronbach's alpha

Hi

I tested a group of 6 parents, who completed a questionnaire, which
contained 15 questions (likert scale response format).

This was part of a pilot study. I used a previously validated measure;
however, the words 'mental health service users' was changed to 'parents
of children with ADHD', to suit my sample. Participants also provided
qualitative data regarding the usefulness of the scale for use with
parents of children with ADHD.

I ran the analysis to determine the Cronbach's alpha for the modified
scale (which will be used in the main study).

Cronbach's was .920

BUT>>>>this warning message comes up on SPSS:

'The determinant of the covariance matrix is zero or approximately zero.
Statistics based on its inverse matrix cannot be computed and they are
displayed as system missing values.'

and no squared multiple correlations were calculated.


What would have caused this to happen?

Parents felt that a few items needed to be modified for various reasons
(hence their responses to these items were not as expected), could the
problem be related to this?

Can I still use the Cronbach's?


Need help, not a stat guru.

Thanks

=====================
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

=====================
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

[ratkc]

Hi there,

I am facing the same problem.

I already tried to use Guttman and Split Half, but this statement "The determinant of the covariance matrix is zero or approximately zero. Statistics based on its inverse matrix cannot be computed and they are displayed as system missing values." still there.

I have analyzed 73 respondents with 23 variables.

If possible, I do appreciate any advices from anyone.

Can I still use Cronbach's Alpha to state that my survey was reliable?

Thank you so much.

Cheers,
Riduan

 

[Rich Ulrich]

There are several reasons why a determinant can be zero.

In the earlier post, where the N was less than the number
of variables, that was sufficient reason -- the determinant
is zero if the matrix is not "full-rank", since the N is a
maximum of the rank of a covariance matrix.  Rank could be less.

Also, a matrix is not full rank if any variable is a "constant":
that is, zero for some standard deviation.

A matrix is not full rank if any variable is a linear
combination of the other variables....  This happens when
a variable happens to be entered twice, or when the list
of variables includes a set subscores and their total.

>
> I am facing the same problem.
>
> I already tried to use Guttman and Split Half, but this statement "The
> determinant of the covariance matrix is zero or approximately zero.
> Statistics based on its inverse matrix cannot be computed and they are
> displayed as system missing values." still there.
>
> I have analyzed 73 respondents with 23 variables.
>
> If possible, I do appreciate any advices from anyone.
>
> Can I still use Cronbach's Alpha to state that my survey was reliable?

It still does show the internal reliability, yes.

--
Rich Ulrich




=====================
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

[ratkc]

Heaps of thanks.

It's really help me a lot.

Kind Regards,

Riduan

---

From: Rich Ulrich-2[via SPSSX Discussion] <[hidden email]>
To: ratkc <[hidden email]>
Sent: Saturday, 30 April 2011 8:46 AM
Subject: Re: Cronbach's alpha

There are several reasons why a determinant can be zero.

In the earlier post, where the N was less than the number
of variables, that was sufficient reason -- the determinant
is zero if the matrix is not "full-rank", since the N is a
maximum of the rank of a covariance matrix.  Rank could be less.

Also, a matrix is not full rank if any variable is a "constant":
that is, zero for some standard deviation.

A matrix is not full rank if any variable is a linear
combination of the other variables....  This happens when
a variable happens to be entered twice, or when the list
of variables includes a set subscores and their total.

>
> I am facing the same problem.
>
> I already tried to use Guttman and Split Half, but this statement "The
> determinant of the covariance matrix is zero or approximately zero.
> Statistics based on its inverse matrix cannot be computed and they are
> displayed as system missing values." still there.
>
> I have analyzed 73 respondents with 23 variables.
>
> If possible, I do appreciate any advices from anyone.
>
> Can I still use Cronbach's Alpha to state that my survey was reliable?

It still does show the internal reliability, yes.

--
Rich Ulrich




=====================
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

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/Cronbach-s-alpha-tp1089680p4360049.html

To unsubscribe from Cronbach's alpha, click here.