Cronbach's and Kuder-Richardson Alpha: How Low?

[Chao Yawo]

Hello,

I am trying to develop some summed scales using "Demographic and Health Surveys". I know that alphas greater than .70 are often acceptable. But is this cut-off dependent on field?  Some have suggested that alphas lower than .60 are acceptable in the social sciences and humanities.

Is there any credence to this? Any sources to support this claim? I heard Hatcher L(1994) makes such a claim in his "Step-by-Step Approach to Using SAS System in Factor Analysis and Structural Equation Modelling", but I am yet to get the book from ILL.

Hoping for some clarification...

thanks Yy

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

[statisticsdoc]

Chao,

When alpha falls below .7. more than half of the variance in scale scores is either unique item variance or error.   This is not good in any discipline

The alpha level that is acceptable often depends on the content of the test and the use to which it is being used.  For a measure of cognitive abiliities that is being used to make an irreversible decision about an individual (such as academic admissions), alpha levels higher than .9 would be expected, as you really do not want such a decision to be made on the basis of random measurement error if you can help it.  On the other hand, for a measure of attitudes that will be used only to make aggregate comparisons between groups, then a lower alpha (.8 and perhaps .7) might be accepted.

HTH,

Steve Brand




---- Chao Yawo <[hidden email]> wrote:

> Hello,
>
> I am trying to develop some summed scales using "Demographic and Health Surveys". I know that alphas greater than .70 are often acceptable. But is this cut-off dependent on field?  Some have suggested that alphas lower than .60 are acceptable in the social sciences and humanities.
>
> Is there any credence to this? Any sources to support this claim? I heard Hatcher L(1994) makes such a claim in his "Step-by-Step Approach to Using SAS System in Factor Analysis and Structural Equation Modelling", but I am yet to get the book from ILL.
>
> Hoping for some clarification...
>
> thanks Yy
>
> =====================
> 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

--
For personalized and experienced consulting in statistics and research design, visit www.statisticsdoc.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

[Chao Yawo]

Thanks for all your responses regarding my question. I agree with all your suggestions, and I will look deeper into my data for other indicators of internal consistency, and take a decision guided by my disciplinary norms and expectations.

Thanks and have a happy Easter !

rgds, CY

--- On Fri, 4/10/09, Chao Yawo <[hidden email]> wrote:

> From: Chao Yawo <[hidden email]>
> Subject: Cronbach's and Kuder-Richardson Alpha:  How Low?
> To: [hidden email], "SAS-L" <[hidden email]>
> Date: Friday, April 10, 2009, 1:22 PM
> Hello,
>
> I am trying to develop some summed scales using
> "Demographic and Health Surveys". I know that alphas greater
> than .70 are often acceptable. But is this cut-off dependent
> on field?  Some have suggested that alphas lower than
> .60 are acceptable in the social sciences and
> humanities.
>
> Is there any credence to this? Any sources to support this
> claim? I heard Hatcher L(1994) makes such a claim in his
> "Step-by-Step Approach to Using SAS System in Factor
> Analysis and Structural Equation Modelling", but I am yet to
> get the book from ILL.
>
> Hoping for some clarification...
>
> thanks Yy
>
>
>
>

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