can apply reliability analysis to ordinary or nominal data ?

[along zeng]

Hi all,
     the common result of reliability analysis is correlation
coefficient,so I think it can't calculate the coefficient of ordinary
or nominal variables;second ,In spss software,this analysis is under
the "scale" menu,does it mean that this analysis only be used to
scale variales?
    what about your opinion?
    TKS!
yours !
                                 xh.along

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

[Ruben Geert van den Berg]

 
I think that reliability analysis in SPSS is usually used in order to check whether a number of items can make up a scale by taking the mean/sum over them. With optimal scaling methods you could assess to what extent a single factor model adequately represents a number of nominal/ordinal variables (which doesn't necessarily mean their reliability is good). 
 
However, the original variables don't have a common unit of measurement (by definition; they would otherwise be scale variables).
Therefore, I guess you cannot construct a scale with nominal/ordinal variables and therefore reliability analysis is useless for such data. Perhaps you could (very!) tentatively try and build a scale of optimally quantified categorical variables... I've never heard of such an attempt, it would be pretty bold I guess.
 
HTH,

Ruben van den Berg
 



 

> Date: Fri, 31 Jul 2009 16:31:49 +0800
> From: [hidden email]
> Subject: can apply reliability analysis to ordinary or nominal data ?
> To: [hidden email]
>
> Hi all,
> the common result of reliability analysis is correlation
> coefficient,so I think it can't calculate the coefficient of ordinary
> or nominal variables;second ,In spss software,this analysis is under
> the "scale" menu,does it mean that this analysis only be used to
> scale variales?
> what about your opinion?
> TKS!
> yours !
> xh.along
>
> =====================
> 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

---

Express yourself instantly with MSN Messenger! MSN Messenger

[Art Kendall]

Please give more detail about your situation.

Are there items designed to make up a scale?

Are you looking for internal consistency reliability of a summative scale?

Are you looking for intercoder or interrater reliability?


please give explicit details of your response scale. By ordinary do you
mean ordinal?  Are your nominal level variables dichotomous?


Art Kendall
Social Research Consultants

>      the common result of reliability analysis is correlation
> coefficient,so I think it can't calculate the coefficient of ordinary
> or nominal variables;second ,In spss software,this analysis is under
> the "scale" menu,does it mean that this analysis only be used to
> scale variales?
>     what about your opinion?
>     TKS!
> yours !
>                                  xh.along
>
> =====================
> 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