[IPAC-List] Reliability for T&Es

[dputka at humrro.org]

Bryan,

Got it.

Are evaluators providing ratings of the applicant's responses?  Or is there
some mechanical scoring key in place that combines information reflected in
the applicant's responses?  In other words, is the concern here with the
ability of raters to reliably rate the information provided by applicants,
or the quality of information provided by applicants?

If the former, some index of interrater reliability (or interrater
agreement, depending on how the scores are being used), may be in order.
One caution I'll offer here is that if you are going with an index of
interrater reliability or agreement, the type of index you should use will
heavily depend on your measurement design (e.g., do all raters rate all
applicants?; is each applicant rated by a unique set of raters?, or is it
more of a mix-  raters vary in their degree of overlap across applicants).

Again, I hate to shamelessly plug my own work, but see:  Putka, D.J., Le,
H., McCloy, R.A., & Diaz, T. (2008). Ill-structured measurement designs in
organizational research: Implications for estimating interrater
reliability. Journal of Applied Psychology, 93, 959-981.

Dan



From:	"Bryan Baldwin" <Bryan.Baldwin at doj.ca.gov>
To:	<dputka at humrro.org>
Cc:	<ipac-list at ipacweb.org>
Date:	06/22/2010 11:28 AM
Subject:	Re: [IPAC-List] Reliability for T&Es



Thanks for the responses.  To be honest, I am the beginning of
investigating this issue, and the type of T&Es I'm thinking of are
primarily task-based.  For those of us in a civil service system where
banding is mandatory, the increasing use of T&Es begs the question of how
these bands should be formed.  I suspect in many cases it's done simply by
dividing the raw spread by the number of ranks.  To speak to Dennis'
question, I have also suspected the bands should be quite large,
particularly at the top given inflation tendencies.  Of course the pass
point obviously plays an issue here, and in my experience SMEs tend to want
to set a higher pass point than is really necessary, which plays into the
size of the bands.

Dan, I appreciate your questions.  As for how I'm defining error, I'm
slightly more concerned about internal consistency than I am about
test-retest, but in reality most concerned about the accuracy of ratings.
What makes this even more complicated is the quality of the T&E items and
scales themselves vary dramatically, much more so than your typical m-c
item.

To answer your second question, a single overall score is being generated.
I have zero faith in the ability of these exams to measure separate
distinct constructs.

BB


>>> <dputka at humrro.org> 6/22/2010 7:06 AM >>>

Hi Bryan,

A few questions....

1. How are you defining error for your T&E measure?

a. As inconsistency across items?   In other words, do you want to draw
inferences with regard to the consistency in your T&E scores if they were
based on different sampling of items?

b. As inconsistency across occasions?  In other words, do you want to draw
inferences with regard to the consistency in your T&E scores if respondents
completed your measure on a different occasion?

c. Are you concerned about both types of inconsistency referenced above?


2. What is the substantive nature of your T&E measure?

a. Are you trying to assess multiple, distinct constructs with your T&E
measure (e.g., each construct indicated by a subset of items comprising the
T&E measure)?

b. Or, do you simply have an overarching, heterogenous measure of T&E?

Depending on how you define error (which is a function of the inferences
you want to make with regard to the consistency of your scores), and the
substantive nature of your T&E measure, it will dictate the structure of
the reliability coefficient that is most appropriate for your situation.

For a general discussion of this, see:
Putka, D. J. & Sackett, P .R. (2010).  Reliability and validity. In J.L.
Farr, & N.T. Tippins (Eds.). Handbook of Employee Selection (pp. 9-49). New
York: Routledge.


Hope this helps,

Dan


*******************************************************************
Dan J. Putka, Ph.D.
Principal Staff Scientist
Personnel Selection & Development Program
Human Resources Research Organization (HumRRO)
66 Canal Center Plaza, Suite 700
Alexandria, VA  22314-1591
703-706-5640 (o)
703-548-5574 (f)
http://www.humrro.org
*******************************************************************




From:"Bryan Baldwin" <Bryan.Baldwin at doj.ca.gov>
To:ipac-list at ipacweb.org
Date:06/21/2010 07:05 PM
Subject:[IPAC-List] Reliability for T&Es
Sent by:ipac-list-bounces at ipacweb.org



Can anyone point me to a source for how to compute reliability for T&Es
(assuming it can be done)?  Ultimate goal is to use this to create bands.

Thanks-

Bryan Baldwin
Staff Services Manager II
California Department of Justice
Division of Administrative Support
Personnel Programs
(916) 322-5446


CONFIDENTIALITY NOTICE: This communication with its contents may contain
confidential and/or legally privileged information. It is solely for the
use of the intended recipient(s). Unauthorized interception, review, use or
disclosure is prohibited and may violate applicable laws including the
Electronic Communications Privacy Act. If you are not the intended
recipient, please contact the sender and destroy all copies of the
communication.
_______________________________________________________
IPAC-List
IPAC-List at ipacweb.org
http://www.ipacweb.org/mailman/listinfo/ipac-list





CONFIDENTIALITY NOTICE: This communication with its contents may contain
confidential and/or legally privileged information. It is solely for the
use of the intended recipient(s). Unauthorized interception, review, use or
disclosure is prohibited and may violate applicable laws including the
Electronic Communications Privacy Act. If you are not the intended
recipient, please contact the sender and destroy all copies of the
communication.