Cronbach's Alpha

[Karen Cacciattolo]

Hello all, I have starting using SPSS, and was wondering whether
someone can help me out on the above mentioned subject.  I have a set
of 10 statements and the Alpha coefficient resulted in 0.5 something,
which of course is not the acceptable value for internal reliability.
I then ran descriptives to check for any errors in the data and
noticed that the maximum number of the scale of one of the statements
is 3, when it is supposed to be 4 (the scale for all statements is 1
to 4).  Does anyone know how to correct this error?
Many thanks,
Karen Cacciattolo
University of Malta

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

[Art Kendall]

Assuming that you are referring to the response scale for an item, it is possible that nobody used the value 4 for that item.  However that in itself would not be the cause of the low alpha.

Some things to check.  Are items properly reflected?  Having non-trivial negative correlations among items would be a hint.
Another would be that the SMC squared multiple correlations are high but it corrected item-total correlations are low.

Were the ten items written to reflect a single construct? What does a Factor analysis with PAF and Varimax rotation show?

Are there items which show an increase in alpha in the "alpha if item deleted" column of the output?

Do you have a lot of non-response?

Do you have a substantial number of cases?

Were the items written for this analysis or are they from a previously developed scale?

Art Kendall
Social Research Consultants


Karen Cacciattolo wrote:

Hello all, I have starting using SPSS, and was wondering whether
someone can help me out on the above mentioned subject.  I have a set
of 10 statements and the Alpha coefficient resulted in 0.5 something,
which of course is not the acceptable value for internal reliability.
I then ran descriptives to check for any errors in the data and
noticed that the maximum number of the scale of one of the statements
is 3, when it is supposed to be 4 (the scale for all statements is 1
to 4).  Does anyone know how to correct this error?
Many thanks,
Karen Cacciattolo
University of Malta

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

[Karen Cacciattolo]

Many thanks for this .. very helpful indeed!  I will go through your
questions and let you know the outcome.
Best regards,
Karen

Quoting Art Kendall <[hidden email]>:

> Assuming that you are referring to the *response *scale for an item, it
> is possible that nobody used the value 4 for that item.  However that in
> itself would not be the cause of the low alpha.
>
> Some things to check.  Are items properly reflected?  Having non-trivial
> negative correlations among items would be a hint.
> Another would be that the SMC squared multiple correlations are high but
> it corrected item-total correlations are low.
>
> Were the ten items written to reflect a single construct? What does a
> Factor analysis with PAF and Varimax rotation show?
>
> Are there items which show an increase in alpha in the "alpha if item
> deleted" column of the output?
>
> Do you have a lot of non-response?
>
> Do you have a substantial number of cases?
>
> Were the items written for this analysis or are they from a previously
> developed scale?
>
> Art Kendall
> Social Research Consultants
>
>
> Karen Cacciattolo wrote:
>> Hello all, I have starting using SPSS, and was wondering whether
>> someone can help me out on the above mentioned subject.  I have a set
>> of 10 statements and the Alpha coefficient resulted in 0.5 something,
>> which of course is not the acceptable value for internal reliability.
>> I then ran descriptives to check for any errors in the data and
>> noticed that the maximum number of the scale of one of the statements
>> is 3, when it is supposed to be 4 (the scale for all statements is 1
>> to 4).  Does anyone know how to correct this error?
>> Many thanks,
>> Karen Cacciattolo
>> University of Malta
>>
>> =====================
>> 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

[J P-6]

Dear List,

 

I am hoping that someone can tell me where I can find the formula for computing the individual prediction interval that can be produced in SPSS using the regression command. In the syntax below, it is the / SAVE ICIN subcommand. I need to generate a prediciton equation and use the formula to give examples of the prediction interval for specific sets of values. I cannot use the regression syntax because the final regression equation is the average of 200 bootstraps.

REGRESSION

/MISSING LISTWISE

/STATISTICS COEFF OUTS R ANOVA

/CRITERIA=PIN(.05) POUT(.10) CIN(95)

/NOORIGIN

/DEPENDENT SATT

/METHOD=ENTER SATM SATV

/SAVE ICIN .

 

Any advice is helpful. Thanks,

John

 

 

[vlad simion]

Hi John,

search the spss site for algorithms used in computation and you will find it

http://support.spss.com/Tech/default.asp

hth,
vlad

On Tue, Feb 3, 2009 at 6:17 PM, J P <[hidden email]> wrote:

Dear List,

 

I am hoping that someone can tell me where I can find the formula for computing the individual prediction interval that can be produced in SPSS using the regression command. In the syntax below, it is the / SAVE ICIN subcommand. I need to generate a prediciton equation and use the formula to give examples of the prediction interval for specific sets of values. I cannot use the regression syntax because the final regression equation is the average of 200 bootstraps.

REGRESSION

/MISSING LISTWISE

/STATISTICS COEFF OUTS R ANOVA

/CRITERIA=PIN(.05) POUT(.10) CIN(95)

/NOORIGIN

/DEPENDENT SATT

/METHOD=ENTER SATM SATV

/SAVE ICIN .

 

Any advice is helpful. Thanks,

John

 

 

[Drake, Lorin]

Hello:

 

I have two variables that were separate questions on the survey:

·         Q605 Tenure with the company (one year or less, >1 but less than four, 4-10 years, 10+ years)

·         Q655 Tenure in the industry (one year or less, >1 but less than four, 4-10 years, 10+ years)

 

How do I combine these two variables into three new categories:

·         New to company/New to industry (< 4 years on both)

·         New to company/Experienced in industry (<4 years on first, >4 years on 2^nd)

·         Experienced (>4 years on both)

I would love some help!

Thanks in advance.

 

Sincerely,

Lorin

 

[Art Kendall]

see if something like this untested syntax is what you want.


recode q605 q655 (lo thru 1 =1)(1 thru 3.99 =2) (4 thru 10 =3)(10 thru hi=4)(else=copy) into incompany inindustry.
value labels incompany inindustry 1 'one year or less' 2 '>1 but less than four' 3'4-10 years' 4 '10+ years'.
crosstabs tables = incompany by inindustry.
do   if any(incompany,1,2) and any(inindustry,1,2).
compute experience =1.
else if any(incompany,1,2) and any(inindustry,3,4).
compute experience =2.
else if any(incompany,3,4) and any(inindustry,3,4).
compute experience =3.
else.
compute experience =4.
end if.
value labels experience
  1 'New to company/New to industry (< 4 years on both)'
  2 'New to company/Experienced in industry (<4 years on first, >4 years on 2nd)'

  3 'Experienced (>4 years on both)'
  4 'fell through cracks'.

Art Kendall
Social Research Consultants

Drake, Lorin wrote:

Hello:

 

I have two variables that were separate questions on the survey:

·         Q605 Tenure with the company (one year or less, >1 but less than four, 4-10 years, 10+ years)

·         Q655 Tenure in the industry (one year or less, >1 but less than four, 4-10 years, 10+ years)

 

How do I combine these two variables into three new categories:

·         New to company/New to industry (< 4 years on both)

·         New to company/Experienced in industry (<4 years on first, >4 years on 2^nd)

·         Experienced (>4 years on both)

I would love some help!

Thanks in advance.

 

Sincerely,

Lorin

 

[Luca Meyer-3]

Supposing that both variables are string the NEW variable will contain the classification you seek:

STRING NEW (A100).
IF (Q605="one year or less" OR Q605=">1 but less than four") AND (Q655="one year or less" OR Q655=">1 but less than four")
NEW="New to company/New to industry (< 4 years on both)".
IF (Q605="one year or less" OR Q605=">1 but less than four") AND (Q655="4-10 years" OR Q655="10+ years")
NEW="New to company/Experienced in industry (<4 years on first, >4 years on 2nd)".
IF (Q605="4-10 years" OR Q605="10+ years") AND (Q655="4-10 years" OR Q655="10+ years")
NEW="Experienced (>4 years on both)".
EXE.

HTH,

Luca
www.lucameyer.com

2009/2/3 Drake, Lorin <[hidden email]>

Hello:

 

I have two variables that were separate questions on the survey:

·         Q605 Tenure with the company (one year or less, >1 but less than four, 4-10 years, 10+ years)

·         Q655 Tenure in the industry (one year or less, >1 but less than four, 4-10 years, 10+ years)

 

How do I combine these two variables into three new categories:

·         New to company/New to industry (< 4 years on both)

·         New to company/Experienced in industry (<4 years on first, >4 years on 2^nd)

·         Experienced (>4 years on both)

I would love some help!

Thanks in advance.

 

Sincerely,

Lorin

 

[Hesham Abdel-Meguid Abdalla]

Hello Karen

you can earse some of 10 items to increase your Alpha coefficient. this can be done using the option "scale if item deleted" under option "statistics". If you use this option, the results will inform you which item has to delete to increase Alpha.

best regards

Hesham Abdel-Meguid Abdalla
Statistics Dept.
Assiut University

On Tue, Feb 3, 2009 at 5:05 PM, Karen Cacciattolo <[hidden email]> wrote:

Hello all, I have starting using SPSS, and was wondering whether
someone can help me out on the above mentioned subject.  I have a set
of 10 statements and the Alpha coefficient resulted in 0.5 something,
which of course is not the acceptable value for internal reliability.
I then ran descriptives to check for any errors in the data and
noticed that the maximum number of the scale of one of the statements
is 3, when it is supposed to be 4 (the scale for all statements is 1
to 4).  Does anyone know how to correct this error?
Many thanks,
Karen Cacciattolo
University of Malta

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

--

[Karen Cacciattolo]

Hello Art and Hesham,

I deleted some of the items which had negative values, the Alpha improved a little bit but it is still low.  I had reversed the scores just in case as well, but the results came out the same!

Probably there is simply no correlation between the items and these are measuring something different.  I will now try to see what the Factor analysis with PAF and Varimax show. 

Regards,

Karen

----- Original Message -----

From: [hidden email]

To: [hidden email]

Cc: [hidden email]

Sent: Tuesday, February 03, 2009 7:10 PM

Subject: Re: Cronbach's Alpha

Hello Karen

you can earse some of 10 items to increase your Alpha coefficient. this can be done using the option "scale if item deleted" under option "statistics". If you use this option, the results will inform you which item has to delete to increase Alpha.

best regards

Hesham Abdel-Meguid Abdalla
Statistics Dept.
Assiut University

On Tue, Feb 3, 2009 at 5:05 PM, Karen Cacciattolo <[hidden email]> wrote:

Hello all, I have starting using SPSS, and was wondering whether
someone can help me out on the above mentioned subject.  I have a set
of 10 statements and the Alpha coefficient resulted in 0.5 something,
which of course is not the acceptable value for internal reliability.
I then ran descriptives to check for any errors in the data and
noticed that the maximum number of the scale of one of the statements
is 3, when it is supposed to be 4 (the scale for all statements is 1
to 4).  Does anyone know how to correct this error?
Many thanks,
Karen Cacciattolo
University of Malta

=====================
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]

If you expect the variables to be related, don’t use varimax, use an oblique rotation!

 

Paul R. Swank, Ph.D

Professor and Director of Research

Children's Learning Institute

University of Texas Health Science Center

Houston, TX 77038

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Karen
Sent: Wednesday, February 04, 2009 1:21 PM
To: [hidden email]
Subject: Re: Cronbach's Alpha

 

Hello Art and Hesham,

I deleted some of the items which had negative values, the Alpha improved a little bit but it is still low.  I had reversed the scores just in case as well, but the results came out the same!

Probably there is simply no correlation between the items and these are measuring something different.  I will now try to see what the Factor analysis with PAF and Varimax show. 

Regards,

Karen

----- Original Message -----

From: [hidden email]

To: [hidden email]

Cc: [hidden email]

Sent: Tuesday, February 03, 2009 7:10 PM

Subject: Re: Cronbach's Alpha

 

Hello Karen

you can earse some of 10 items to increase your Alpha coefficient. this can be done using the option "scale if item deleted" under option "statistics". If you use this option, the results will inform you which item has to delete to increase Alpha.

best regards

Hesham Abdel-Meguid Abdalla
Statistics Dept.
Assiut University

On Tue, Feb 3, 2009 at 5:05 PM, Karen Cacciattolo <[hidden email]> wrote:

Hello all, I have starting using SPSS, and was wondering whether
someone can help me out on the above mentioned subject.  I have a set
of 10 statements and the Alpha coefficient resulted in 0.5 something,
which of course is not the acceptable value for internal reliability.
I then ran descriptives to check for any errors in the data and
noticed that the maximum number of the scale of one of the statements
is 3, when it is supposed to be 4 (the scale for all statements is 1
to 4).  Does anyone know how to correct this error?
Many thanks,
Karen Cacciattolo
University of Malta

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

--

[Art Kendall]

If items have several sizable negative interitem correlations read their content. It may be that they should be reflected rather than deleted.  Use PAF because in scale construction you are interested in finding items that are weak measures of a construct that  you hope will combine into a stronger measure of the construct.

When you do the PAF interpret the meaning of the factor and create a summative score bases on items that load cleanly on a factor.  If some items have negative and positive loadings on a factor interpret and score based on a reversal of meaning.  Be sure to use an item on only one scale.  In scale construction divergent validity is aided by the use of varimax rotation. 

Art Kendall
Social Research Consultants

Karen wrote:

Hello Art and Hesham,

I deleted some of the items which had negative values, the Alpha improved a little bit but it is still low.  I had reversed the scores just in case as well, but the results came out the same!

Probably there is simply no correlation between the items and these are measuring something different.  I will now try to see what the Factor analysis with PAF and Varimax show. 

Regards,

Karen

----- Original Message -----

From: [hidden email]

To: [hidden email]

Cc: [hidden email]

Sent: Tuesday, February 03, 2009 7:10 PM

Subject: Re: Cronbach's Alpha

Hello Karen

you can earse some of 10 items to increase your Alpha coefficient. this can be done using the option "scale if item deleted" under option "statistics". If you use this option, the results will inform you which item has to delete to increase Alpha.

best regards

Hesham Abdel-Meguid Abdalla
Statistics Dept.
Assiut University

On Tue, Feb 3, 2009 at 5:05 PM, Karen Cacciattolo <[hidden email]> wrote:

Hello all, I have starting using SPSS, and was wondering whether
someone can help me out on the above mentioned subject.  I have a set
of 10 statements and the Alpha coefficient resulted in 0.5 something,
which of course is not the acceptable value for internal reliability.
I then ran descriptives to check for any errors in the data and
noticed that the maximum number of the scale of one of the statements
is 3, when it is supposed to be 4 (the scale for all statements is 1
to 4).  Does anyone know how to correct this error?
Many thanks,
Karen Cacciattolo
University of Malta

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

--

[Bob Schacht-3]

At 12:03 PM 2/4/2009, Swank, Paul R wrote:
>If you expect the variables to be related, don't use varimax, use an
>oblique rotation!


What oblique rotation options are available in SPSS?

Bob Schacht

Robert M. Schacht, Ph.D. <[hidden email]>
Pacific Basin Rehabilitation Research & Training Center
1268 Young Street, Suite #204
Research Center, University of Hawaii
Honolulu, HI 96814

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