Appropriate to use Cronbach Alpha on a Yes/No scale

[Mark Webb-5]

I'm attempting this and getting very low Alphas.

Is it advisable to use the measure with a 2 point scale?

[Swank, Paul R]

Certainly. The original Kuder-Richardson 20 formula is the dichotomous version of Cronbach’s alpha. You need to look at the means and variances of the items as well and their correlation with the total score. That will likely indicate where the problem is.

 

Dr. Paul R. Swank,

Children's Learning Institute

Professor, Department of Pediatrics, Medical School

Adjunct Professor, School of Public Health

University of Texas Health Science Center-Houston

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mark Webb
Sent: Thursday, November 03, 2011 9:52 AM
To: [hidden email]
Subject: Appropriate to use Cronbach Alpha on a Yes/No scale

 

I'm attempting this and getting very low Alphas.

Is it advisable to use the measure with a 2 point scale?

--

Mark Webb

Line [phone]

Cell [phone] [Poor reception]

Fax  [phone]

Skype       tomarkwebb

Email       [hidden email]

Client ftp  http://targetlinkresearch.co.za/cftp/

[statisticsdoc]

Mark,
You will want to use the Kuder-Richardson 20 formula for dichotomous scales (alpha is a generalization of the KR 20). If you are getting low reliability, look at the item total correlations. Items may have low item-total correlations for a variety of reasons. The item might happen to have very little variability in your sample (so perhaps you need a more diverse sample). The item may have been mistakenly scored in the wrong direction (it happens sometimes). The item might not have a very strong relationship with the construct under investigation (dropping the item will increase internal consistency). If your internal consistency is so low, perhaps you should consider the possibility that the items measure multiple dimensions (which would lead us to a discussion of factor analysis with dichotomous items).
HTH
Stephen Brand, Ph.D.
www.StatisticsDoc.com

www.StatisticsDoc.com

---

From: Mark Webb <[hidden email]>

Sender: "SPSSX(r) Discussion" <[hidden email]>

Date: Thu, 3 Nov 2011 16:51:52 +0200

To: <[hidden email]>

ReplyTo: Mark Webb <[hidden email]>

Subject: Appropriate to use Cronbach Alpha on a Yes/No scale

I'm attempting this and getting very low Alphas.

Is it advisable to use the measure with a 2 point scale?

[Anthony Babinec]

How many items are you summing? The value of coefficient

alpha is related to the average item covariance/correlation and

the number of items. Also, make sure the covariance/correlations

are in the same direction.

 

 

Tony Babinec

[hidden email]

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mark Webb
Sent: Thursday, November 03, 2011 9:52 AM
To: [hidden email]
Subject: Appropriate to use Cronbach Alpha on a Yes/No scale

 

I'm attempting this and getting very low Alphas.

Is it advisable to use the measure with a 2 point scale?

--

Mark Webb

Line [phone]

Cell [phone] [Poor reception]

Fax  [phone]

Skype       tomarkwebb

Email       [hidden email]

Client ftp  http://targetlinkresearch.co.za/cftp/

[Rich Ulrich]

I will expand on Tony's comment -

One version of alpha is computed directly as a function of
average r  and K (number of items), and nothing else.  (That
one assumes that variances are equal, which is not always assumed.)
 - To reach a given alpha, you need 3 or 4 or more times the number
of items when items are dichotomies, not short scales.
- You certainly want to assure that your items are all scored
"in the same direction" in terms of the scale's latent dimension
since the program makes that assumption.
 
And I can expand on a comment in another post:  every measure
of reliability (alpha, ICC, r, whatever) reflects on both the quality of
the scale and the variation existing in the sample.  A sample with
lower variation on the latent dimension will always yield a smaller
measure for its reliability, other sources of error being equal.

--
Rich Ulrich

---

Date: Thu, 3 Nov 2011 11:33:46 -0500
From: [hidden email]
Subject: Re: Appropriate to use Cronbach Alpha on a Yes/No scale
To: [hidden email]

How many items are you summing? The value of coefficient

alpha is related to the average item covariance/correlation and

the number of items. Also, make sure the covariance/correlations

are in the same direction.

 

 

Tony Babinec

[hidden email]

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mark Webb
Sent: Thursday, November 03, 2011 9:52 AM
To: [hidden email]
Subject: Appropriate to use Cronbach Alpha on a Yes/No scale

 

I'm attempting this and getting very low Alphas.

Is it advisable to use the measure with a 2 point scale?