comparing two values of cronbach's alpha

[Hannah Klaas]

Hellow everybody,
I have got an urgent question concerning reliabilty and cronbach's alpha
and would be very pleased if anyone could help!
This is my problem: I have got two reliabilties (e.g. alpha= 0, 48 and
alpha= 0,57) and I want to compare them, that is, I want to show that the
latter one is significantly higher than the first one. What do I have to
do? I am not sure, but I think from these two values you cannot just
say "the second one is much more reliable than the first one" and quantify
the "more". You cannot compare two values of cronbach's alpha just like
that, can you?
I hope there is someone who can help me,
thanks a lot,
Hannah

[Alexander J. Shackman-2]

http://psyphz.psych.wisc.edu/%7Eshackman/mediation_moderation_resources.htm#References_for_Testing_for_Significant_D

relevant articles, spss syntax from andrew hayes, and an .xls-base
calculator

hth, ajs

On 4/14/07, Hannah Klaas <[hidden email]> wrote:

>
> Hellow everybody,
> I have got an urgent question concerning reliabilty and cronbach's alpha
> and would be very pleased if anyone could help!
> This is my problem: I have got two reliabilties (e.g. alpha= 0, 48 and
> alpha= 0,57) and I want to compare them, that is, I want to show that the
> latter one is significantly higher than the first one. What do I have to
> do? I am not sure, but I think from these two values you cannot just
> say "the second one is much more reliable than the first one" and quantify
> the "more". You cannot compare two values of cronbach's alpha just like
> that, can you?
> I hope there is someone who can help me,
> thanks a lot,
> Hannah
>

--
Alexander J. Shackman
Laboratory for Affective Neuroscience
Waisman Laboratory for Brain Imaging & Behavior
University of Wisconsin-Madison
1202 West Johnson Street
Madison, Wisconsin 53706

Telephone: [phone]
FAX: [phone]
EMAIL: [hidden email]
http://psyphz.psych.wisc.edu/~shackman

[Baker, Harley]

Hi Hannah,

Without knowing more about your particular situation, I can't give many specifics. However, in general, it is possible to test the significance of the differences among coefficients alpha. Leonard Feldt has published a number of articles over the years with formulas (and many with examples) that can be applied to different situations (e.g., both alphas based on same examinees, alphas derived from different examinees, small sample alphas, alphas based on different test lengths). My recommendation would be to look in the psychology literature (he publishes mostly in applied measurement journals) for the one that fits your particular situation.

Harley


Dr. Harley Baker
Associate Professor and Chair, Psychology Program
Chief Assessment Officer for Academic Affairs
California State University Channel Islands
One University Drive
Camarillo, CA 93012

805.437.8997 (p)
805.437.8951 (f)

[hidden email] <mailto:[hidden email]>

________________________________

From: SPSSX(r) Discussion on behalf of Hannah Klaas
Sent: Sat 4/14/2007 4:00 AM
To: [hidden email]
Subject: comparing two values of cronbach's alpha



Hellow everybody,
I have got an urgent question concerning reliabilty and cronbach's alpha
and would be very pleased if anyone could help!
This is my problem: I have got two reliabilties (e.g. alpha= 0, 48 and
alpha= 0,57) and I want to compare them, that is, I want to show that the
latter one is significantly higher than the first one. What do I have to
do? I am not sure, but I think from these two values you cannot just
say "the second one is much more reliable than the first one" and quantify
the "more". You cannot compare two values of cronbach's alpha just like
that, can you?
I hope there is someone who can help me,
thanks a lot,
Hannah

[Roberts, Michael]

Hi,

Does anyone in here have any idea of what is causing the following error
to be generated with SPSS identified in event properties(?)

My system is a Dell Optiplex Gx620 dual core 3.4ghz, 2.0gb RAM, 210bg
hdd. WIN XP professional (OS); No apparent problems with SPSS other than
occasional aggregation command not executing.(???)

"
mfc71.dll, version 7.10.3077.0, fault address 0x00018bcd;

Application failure spsswin.ext 14.0.1.  321 in mfc71.dll  7.10.30 77.0
at offset 00018bcd. 0a

"

Any pointers would be very much appreciated.

TIA

Mike

[Susan Mikulich]

Hi,
I have data consisting of about 500 families of adolescent twins and a
few triplets.  I'd like to randomly select one adolescent from each
family.  I know how to use SPSS to randomly assign subjects or randomly
select a percentage of a sample but not to randomly select x cases
(adolescent) from each subset (family).  Can anyone help me?
Thanks,
Susan

Susan Mikulich-Gilbertson, PhD.
Assistant Professor: Joint Appointment
Depts of Psychiatry and Preventive Medicine/Biometrics
Director of Biostatistics, Division of Substance Dependence
University of Colorado Health Sciences Center
Phone: [phone]
Fax: [phone]
E-mail: [hidden email]

[Maguin, Eugene]

Susan,

Here is one way to do this. I won't say this is the only way or even the
shortest way. I'll assume you have a distinct family id number and child id
number.

*  assign a random number to each record.
Compute rn=uniform(1).
*  sort cases by family id and rn. Choose the first case within the family
id.
Sort cases by familyid rn.

Compute pick=0.
Do if ($casenum eq 1).  /* this for the first family only.
+  compute pick=1.
Else.
+  if (familyid ne lag(familyid)) pick=1.
End if.

Select if (pick eq 1).

Gene Maguin




I have data consisting of about 500 families of adolescent twins and a
few triplets.  I'd like to randomly select one adolescent from each
family.  I know how to use SPSS to randomly assign subjects or randomly
select a percentage of a sample but not to randomly select x cases
(adolescent) from each subset (family).  Can anyone help me?
Thanks,
Susan