HELP NEEDED WITH FINDING CRONBACH'S ALPHA

[boye_rich]

Hi all,

I conducted a survey  through a questionnaire  using Likert scale from 1 to 5.
I then wanted to find Cronbach´s alpha  to check the reliability  but the result of the cronbach's alpha always came out as -4,00.  I really need help with this and I will appreciate it if any of you guys can assist me with this as I need it for my thesis work.


Regards, boye_rich

[David Marso]

How is it that you ended up with this -4,0 .
Maybe if you provide useful information, someone can provide useful tips?
For one thing, verify the directionality of how the items are scored?
What are the interitem correlations/covariances?
Beyond that, you need to review your data and the assumptions underlying alpha.

boye_rich wrote

Hi all,

I conducted a survey  through a questionnaire  using Likert scale from 1 to 5.
I then wanted to find Cronbach´s alpha  to check the reliability  but the result of the cronbach's alpha always came out as -4,00.  I really need help with this and I will appreciate it if any of you guys can assist me with this as I need it for my thesis work.


Regards, boye_rich

[Mike]

The message David replied to hasn't gotten to SPSSL (at least I didn't
get it)
but some simple answers to the original questions can be found here:

(1) The UCLA website for SPSS which provides one definition for
Cronbach's alpha and some follow-up; see:
http://www.ats.ucla.edu/stat/spss/faq/alpha.html

In the equation "c-bar" is the average covariance but one can also
imagine it as "r-bar", the average Pearson correlation (once upon
a time UCLA used the correlation form of the equation but now
use the covariance form since that is more general and is more
appropriate if one follows up with structural equation modeling (SEM)
rather than factor analysis which the UCLA uses).

(2)  More relevant is the following page which explains why one
might get a negative Cronbach's alpha by Dave Nichols, SPSSL
legend ;-); see:
http://www.ats.ucla.edu/stat/spss/library/negalpha.htm

Simple answer: the mean covariance/correlation is negative.

If you look at the correlation matrix or covariance matrix for the items
used in the reliability, you should not see negative correlations.
Ideally, you want somewhat called "compound symmetry" which means
that the items all have (a) the same variance and (b) correlations
between all items are the same (if the foregoing makes sense to you
then read the paper here:
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&ved=0CDAQFjAB&url=http%3A%2F%2Fwww.cs.unc.edu%2FResearch%2FMIDAG%2Fpubs%2Fpapers%2FPsychometrika04_Kistner.pdf&ei=i4wTU9j6BNKfkQeEiIDABg&usg=AFQjCNGe55vQu5kV6UfgDkxhoZc8iRyk8w&sig2=JqcB6-7-jNiGarf4Im6cvA )
.

But you probably do not know what the population covariance matrix
is (I mean the actual values in the cova matrix, not the concept), so
you're
using the sample covariance/correlation matrix as an estimate in the
calculation
of Cronbach's alpha.  And a sample covariance/correlation matrix could
have negative correlations because of sampling error, you forgot to
reverse the coding on (contrait) items, or some other error.  If you
think that you've done everything right and you think you know your
data well, then I believe the expression that is appropriate is that
you're
SOL.  In other words, your data don't meet the conditions necessary
for appropriate use of Cronbach's alpha.  Indicate how many items
you have and how many negative correlations you among them (if
more than zero, start making a burnt offering to the random number God).

-Mike Palij
New York University
[hidden email]



----- Original Message -----
From: "David Marso" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, March 02, 2014 2:07 PM
Subject: Re: HELP NEEDED WITH FINDING CRONBACH'S ALPHA

> How is it that you ended up with this -4,0 .
> Maybe if you provide useful information, someone can provide useful
> tips?
> For one thing, verify the directionality of how the items are scored?
> What are the interitem correlations/covariances?
> Beyond that, you need to review your data and the assumptions
> underlying
> alpha.
>
>
> boye_rich wrote
>> Hi all,
>>
>> I conducted a survey  through a questionnaire  using Likert scale
>> from 1
>> to 5.
>> I then wanted to find Cronbach´s alpha  to check the reliability  but
>> the
>> result of the cronbach's alpha always came out as -4,00.  I really
>> need
>> help with this and I will appreciate it if any of you guys can assist
>> me
>> with this as I need it for my thesis work.
>>
>>
>> Regards, boye_rich

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

[Bruce Weaver]

In the Nabble forum for SPSSX-L, I see that the original post in this thread "has NOT been accepted by the mailing list yet".  That makes me wonder if boye_rich (the OP) has created a Nabble account, but has not subscribed to the mailing list.  

If you use Nabble to read & post but have not subscribed to the actual mailing list, please see this page:

http://spssx-discussion.1045642.n5.nabble.com/mailing_list/MailingListOptions.jtp?forum=1068821

HTH.

Mike Palij wrote

The message David replied to hasn't gotten to SPSSL (at least I didn't
get it)
but some simple answers to the original questions can be found here:

(1) The UCLA website for SPSS which provides one definition for
Cronbach's alpha and some follow-up; see:
http://www.ats.ucla.edu/stat/spss/faq/alpha.html

In the equation "c-bar" is the average covariance but one can also
imagine it as "r-bar", the average Pearson correlation (once upon
a time UCLA used the correlation form of the equation but now
use the covariance form since that is more general and is more
appropriate if one follows up with structural equation modeling (SEM)
rather than factor analysis which the UCLA uses).

(2)  More relevant is the following page which explains why one
might get a negative Cronbach's alpha by Dave Nichols, SPSSL
legend ;-); see:
http://www.ats.ucla.edu/stat/spss/library/negalpha.htm

Simple answer: the mean covariance/correlation is negative.

If you look at the correlation matrix or covariance matrix for the items
used in the reliability, you should not see negative correlations.
Ideally, you want somewhat called "compound symmetry" which means
that the items all have (a) the same variance and (b) correlations
between all items are the same (if the foregoing makes sense to you
then read the paper here:
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&ved=0CDAQFjAB&url=http%3A%2F%2Fwww.cs.unc.edu%2FResearch%2FMIDAG%2Fpubs%2Fpapers%2FPsychometrika04_Kistner.pdf&ei=i4wTU9j6BNKfkQeEiIDABg&usg=AFQjCNGe55vQu5kV6UfgDkxhoZc8iRyk8w&sig2=JqcB6-7-jNiGarf4Im6cvA )
.

But you probably do not know what the population covariance matrix
is (I mean the actual values in the cova matrix, not the concept), so
you're
using the sample covariance/correlation matrix as an estimate in the
calculation
of Cronbach's alpha.  And a sample covariance/correlation matrix could
have negative correlations because of sampling error, you forgot to
reverse the coding on (contrait) items, or some other error.  If you
think that you've done everything right and you think you know your
data well, then I believe the expression that is appropriate is that
you're
SOL.  In other words, your data don't meet the conditions necessary
for appropriate use of Cronbach's alpha.  Indicate how many items
you have and how many negative correlations you among them (if
more than zero, start making a burnt offering to the random number God).

-Mike Palij
New York University
[hidden email]



----- Original Message -----
From: "David Marso" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, March 02, 2014 2:07 PM
Subject: Re: HELP NEEDED WITH FINDING CRONBACH'S ALPHA


> How is it that you ended up with this -4,0 .
> Maybe if you provide useful information, someone can provide useful
> tips?
> For one thing, verify the directionality of how the items are scored?
> What are the interitem correlations/covariances?
> Beyond that, you need to review your data and the assumptions
> underlying
> alpha.
>
>
> boye_rich wrote
>> Hi all,
>>
>> I conducted a survey  through a questionnaire  using Likert scale
>> from 1
>> to 5.
>> I then wanted to find Cronbach´s alpha  to check the reliability  but
>> the
>> result of the cronbach's alpha always came out as -4,00.  I really
>> need
>> help with this and I will appreciate it if any of you guys can assist
>> me
>> with this as I need it for my thesis work.
>>
>>
>> Regards, boye_rich

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

[Ware, William B]

Are there items that need to be reverse-scored?  That is only way that I've ever seen a negative alpha.

Bill

William B. Ware, Ph.D.� � � �
Professor of Learning Sciences and Psychological Studies
Adjunct Professor, School of Social Work
McMichael Term Professor of Education 2011-2013
Academy of Distinguished Teaching Scholars at UNC-Chapel Hill – Charter Member
CB #3500 - 118 Peabody Hall�
University of North Carolina at Chapel Hill
Chapel Hill, NC� � � �  27599-3500
Office:� �  (919)-962-0132�  (Sharon Powell - 309 Peabody)
Fax:� � � � � � �  (919)-962-1533
Office:�  118 Peabody Hall
EMAIL: [hidden email]


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of David Marso
Sent: Sunday, March 02, 2014 2:07 PM
To: [hidden email]
Subject: Re: HELP NEEDED WITH FINDING CRONBACH'S ALPHA

How is it that you ended up with this -4,0 .
Maybe if you provide useful information, someone can provide useful tips?
For one thing, verify the directionality of how the items are scored?
What are the interitem correlations/covariances?
Beyond that, you need to review your data and the assumptions underlying alpha.


boye_rich wrote

> Hi all,
>
> I conducted a survey  through a questionnaire  using Likert scale from
> 1 to 5.
> I then wanted to find Cronbach´s alpha  to check the reliability  but
> the result of the cronbach's alpha always came out as -4,00.  I really
> need help with this and I will appreciate it if any of you guys can
> assist me with this as I need it for my thesis work.
>
>
> Regards, boye_rich

-----
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me.
---
"Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis."
Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?"
--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/HELP-NEEDED-WITH-FINDING-CRONBACH-S-ALPHA-tp5724680p5724686.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

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

[Rich Ulrich]

How do you see a Cronbach alpha of -4.00?  What good/bad
values are presented for other statistics?  "-4" is not in the
the computational range, is it?

If -4  was reasonable for any other value in the listing,
I would guess that someone was reading the wrong
field.  Since I can't think of any value that would be -4,
I imagine that the copy of SPSS has been corrupted.

Before re-loading, however, I would make sure that the
data are not, somehow, totally strange:  The correlation
matrix is not presented by default, but that is an easy
thing to look at. 

--
Rich Ulrich

> Date: Sun, 2 Mar 2014 11:07:08 -0800

> From: [hidden email]
> Subject: Re: HELP NEEDED WITH FINDING CRONBACH'S ALPHA
> To: [hidden email]
>
> How is it that you ended up with this -4,0 .
> Maybe if you provide useful information, someone can provide useful tips?
> For one thing, verify the directionality of how the items are scored?
> What are the interitem correlations/covariances?
> Beyond that, you need to review your data and the assumptions underlying
> alpha.
>
>
> boye_rich wrote
> > Hi all,
> >
> > I conducted a survey through a questionnaire using Likert scale from 1
> > to 5.
> > I then wanted to find Cronbach´s alpha to check the reliability but the
> > result of the cronbach's alpha always came out as -4,00. I really need
> > help with this and I will appreciate it if any of you guys can assist me
> > with this as I need it for my thesis work.
> >
> >
> > Regards, boye_rich
>
>
>
>
>
> -----
> Please reply to the list and not to my personal email.
> Those desiring my consulting or training services please feel free to email me.
> ---
> "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis."
> Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?"
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/HELP-NEEDED-WITH-FINDING-CRONBACH-S-ALPHA-tp5724680p5724686.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> 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]

It is inconceivable to me that a "corrupted SPSS" could be the issue here.  If modules are corrupted, almost surely that would simply cause a crash or severity 4 or 5 error.  Reinstalling is not going to change anything.


Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
phone: [phone]




From:        Rich Ulrich <[hidden email]>
To:        [hidden email],
Date:        03/02/2014 04:37 PM
Subject:        Re: [SPSSX-L] HELP NEEDED WITH FINDING CRONBACH'S ALPHA
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

How do you see a Cronbach alpha of -4.00?  What good/bad
values are presented for other statistics?  "-4" is not in the
the computational range, is it?

If -4  was reasonable for any other value in the listing,
I would guess that someone was reading the wrong
field.  Since I can't think of any value that would be -4,
I imagine that the copy of SPSS has been corrupted.

Before re-loading, however, I would make sure that the
data are not, somehow, totally strange:  The correlation
matrix is not presented by default, but that is an easy
thing to look at.  

--
Rich Ulrich

> Date: Sun, 2 Mar 2014 11:07:08 -0800
> From: [hidden email]
> Subject: Re: HELP NEEDED WITH FINDING CRONBACH'S ALPHA
> To: [hidden email]
>
> How is it that you ended up with this -4,0 .
> Maybe if you provide useful information, someone can provide useful tips?
> For one thing, verify the directionality of how the items are scored?
> What are the interitem correlations/covariances?
> Beyond that, you need to review your data and the assumptions underlying
> alpha.
>
>
> boye_rich wrote
> > Hi all,
> >
> > I conducted a survey through a questionnaire using Likert scale from 1
> > to 5.
> > I then wanted to find Cronbach´s alpha to check the reliability but the
> > result of the cronbach's alpha always came out as -4,00. I really need
> > help with this and I will appreciate it if any of you guys can assist me
> > with this as I need it for my thesis work.
> >
> >
> > Regards, boye_rich
>
>
>
>
>
> -----
> Please reply to the list and not to my personal email.
> Those desiring my consulting or training services please feel free to email me.
> ---
> "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis."
> Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?"
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/HELP-NEEDED-WITH-FINDING-CRONBACH-S-ALPHA-tp5724680p5724686.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> 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

[Ryan]

Since my previous attempt at posting to the other thread did not work, I figured I'd start a new thread. I suspect others may be having a difficult time responding as well. 

Below a brief note to add to what has been stated already.

Ryan

--

Cronbach's alpha ranges from negative (-) infinity to positive (+) 1. 

Since Cronbach's alpha is intended to estimate reliability (Rxx), which according to Classical Test Theory, is the proportion:

True Score Variance / Total Variance = Rxx

...a negative alpha is indicative of a problem that should be investigated; that is, evaluation of the assumptions of the model upon which Cronbach's alpha is based ought to be carefully examined. Even when the assumptions are met, a large amount of measurement error could result in a negative alpha. 

HTH.

[Art Kendall]

A very common reason to obtain a negative Alpha is that the scoring key is incorrect.

Is the list of items that define the scale correct? Are there items that belong on other scales?

Are items reflected as defined in the scoring key?  (in a factor analysis these items would have had a different sign for that selected for positive scores.) A negative corrected item-total score can be a flag for this situation.


On 3/2/2014 10:04 PM, Ryan Black [via SPSSX Discussion] wrote:

Since my previous attempt at posting to the other thread did not work, I figured I'd start a new thread. I suspect others may be having a difficult time responding as well. 

Below a brief note to add to what has been stated already.

Ryan

--

Cronbach's alpha ranges from negative (-) infinity to positive (+) 1. 

Since Cronbach's alpha is intended to estimate reliability (Rxx), which according to Classical Test Theory, is the proportion:

True Score Variance / Total Variance = Rxx

...a negative alpha is indicative of a problem that should be investigated; that is, evaluation of the assumptions of the model upon which Cronbach's alpha is based ought to be carefully examined. Even when the assumptions are met, a large amount of measurement error could result in a negative alpha. 

HTH.

---

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

http://spssx-discussion.1045642.n5.nabble.com/Negative-Cronbach-s-Alpha-New-Thread-tp5724693.html

To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

[boye_rich]

My taget population is 4 and I chose 10 questions out of the questionnaire. The choice of answer ranges from 1 to 5 (Likert scale). In one question, all the respondents chose 5 on the likert scale and their answers to the  rest of the questions varied. I used SPSS to find the Cronbach,s alpha because I need to state the reliability of the questionnaire results from the respondents.  

I got -4 as the Cronbach's alpha after I entered the data in SPSS.


The second Cronbach's alpha for the other set of population of 4 and 4 answers from the questionnaire also came out as  -4 .

I can upload the questionnaire on this forum if it will help in getting assistance from someone .


Regards, boy_rich

[David Marso]

You have been directed by SEVERAL people to examine the pattern of correlations. Negative correlations would suggest items have been scored in reverse of what they should be.  The question which everyone scored 5 is USELESS since it has NO VARIATION.  Don't upload anything until you have explored the advise people have already provided!
--

boye_rich wrote

My taget population is 4 and I chose 10 questions out of the questionnaire. The choice of answer ranges from 1 to 5 (Likert scale). In one question, all the respondents chose 5 on the likert scale and their answers to the  rest of the questions varied. I used SPSS to find the Cronbach,s alpha because I need to state the reliability of the questionnaire results from the respondents.  

I got -4 as the Cronbach's alpha after I entered the data in SPSS.


The second Cronbach's alpha for the other set of population of 4 and 4 answers from the questionnaire also came out as  -4 .

I can upload the questionnaire on this forum if it will help in getting assistance from someone .


Regards, boy_rich