Calculating Kappa and McNemar Test in SPSS 15

[Eric Benchimol]

I'm hoping someone can help me.  I have administered a questionnaire to patients (Group=1) and their siblings (Group=0).  One of the questions is on a diagnosis.  I have coded the diagnosis=1 if they get the correct answer, diagnosis=0 if they get the incorrect answer. I have 47 questionnaires to each the patient and siblings, and have coded the matched pair based on an ID number.

I am trying to calculate a Kappa for how often the patient & sibling pairs get the answer correct or incorrect.

My data set looks like this:

ID - GROUP - DIAGNOSIS
1  1  1
2  1  0
3  1  1
4  1  1
5  1  0
6  1  0
1  0  1
2  0  0
3  0  1
4  0  1
5  0  0
6  0  0

In this case, there is perfect agreement.  How do I calculate the Kappa (with SD) using SPSS 15?

Also, I'd like to calculate an McNemar Test in order to determine whether the patient (group 1) gets the answer correct more often than their sibling match.  How do I do this in SPSS?

Thank you in advance for the help!

[Marta Garcia-Granero]

Hi Eric

Eric Benchimol wrote:

> I have administered a questionnaire to patients (Group=1) and their
> siblings (Group=0).  One of the questions is on a diagnosis.  I have
> coded the diagnosis=1 if they get the correct answer, diagnosis=0 if
> they get the incorrect answer. I have 47 questionnaires to each the
> patient and siblings, and have coded the matched pair based on an ID
> number.
>
> I am trying to calculate a Kappa for how often the patient & sibling
> pairs get the answer correct or incorrect.
>
> My data set looks like this:
>
> ID - GROUP - DIAGNOSIS
> 1  1  1
> 2  1  0
> 3  1  1
> 4  1  1
> 5  1  0
> 6  1  0
> 1  0  1
> 2  0  0
> 3  0  1
> 4  0  1
> 5  0  0
> 6  0  0

Answer: wrong layout for the questions, your dataset should looklike this.
ID  Sib1 Sib2
1    1    1
2    0    0
3    1    1
4    1    1
5    0    0
6    0    0
>
> In this case, there is perfect agreement.  How do I calculate the
> Kappa (with SD) using SPSS 15?

If you have a perfect fit, answer is kappa=1. (no  need to ask any program).
>
> Also, I'd like to calculate an McNemar Test in order to determine
> whether the patient (group 1) gets the answer correct more often than
> their sibling match.

Perfect agreement -> McNemar p=1.000 (no need to ask SPSS).

>   How do I do this in SPSS?
Assuming the full dataset doesn't show a perfect fit, this is the way:

* Sample dataset (notice ID=7 cases added by me, to break the perfect
fit) *.
DATA LIST LIST/ID GROUP DIAGNOSIS (3 F8).
BEGIN DATA
1  1  1
2  1  0
3  1  1
4  1  1
5  1  0
6  1  0
7  1  1
1  0  1
2  0  0
3  0  1
4  0  1
5  0  0
6  0  0
7  0  0
END DATA.

* Data must be restructured first (this procedure will fail if data show
an absolutely perfect fit) *.
SORT CASES BY ID GROUP .
CASESTOVARS
 /ID = ID
 /INDEX = GROUP
 /GROUPBY = VARIABLE .

* McNemar & Kappa with SPSS *.
CROSSTABS
  /TABLES=DIAGNOSIS.0  BY DIAGNOSIS.1
  /FORMAT= AVALUE TABLES
  /STATISTIC=KAPPA MCNEMAR
  /CELLS= COUNT
  /COUNT ROUND CELL .

HTH,
Marta García-Granero

--
For miscellaneous SPSS related statistical stuff, visit:
http://gjyp.nl/marta/

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

[Ergul, Emel A.]

Dear List:

Again, my question might be seemingly simple even stupid for many of you who
have been working on meta data. Please accept my apologies if it is so.

CASE: I want to validate my data where the data entry were done with few other
people. I have 900 cases and about 300 variables I have the hard copy of values
(correct) and values that had been entered (correct and incorrect). My plan is
to take XX number of people randomly from the set and review the entries.

QUESTION: How can I set up an upper limit for err for evidence of in/validation?
Do you know any good book or online resource on data validation I can read.

I'm grateful as always for any help,
E.

PS: Majority of resources appeared online is for those who wants to set up a
program prior to data entry for incorrect values. Mine are already there.



The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.

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

[Art Kendall]

Since you do not have a large amount of data, I strongly suggest that
you enter it again.  1)Proofread your data dictionary 2)for as many
people you have to re-enter the data create new entry files using the
same dictionary.  3) have people re-enter the data  4)Use ADD files to
put the new set of files together into a second version.  5) Sort the
original entry file and the second entry file in the same order.


Then search the archives for COMPDS  for how to compare data sets.

Art Kendall
Social Research Consultants

Ergul, Emel A. wrote:

> Dear List:
>
> Again, my question might be seemingly simple even stupid for many of you who
> have been working on meta data. Please accept my apologies if it is so.
>
> CASE: I want to validate my data where the data entry were done with few other
> people. I have 900 cases and about 300 variables I have the hard copy of values
> (correct) and values that had been entered (correct and incorrect). My plan is
> to take XX number of people randomly from the set and review the entries.
>
> QUESTION: How can I set up an upper limit for err for evidence of in/validation?
> Do you know any good book or online resource on data validation I can read.
>
> I'm grateful as always for any help,
> E.
>
> PS: Majority of resources appeared online is for those who wants to set up a
> program prior to data entry for incorrect values. Mine are already there.
>
>
>
> The information in this e-mail is intended only for the person to whom it is
> addressed. If you believe this e-mail was sent to you in error and the e-mail
> contains patient information, please contact the Partners Compliance HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to you in error
> but does not contain patient information, please contact the sender and properly
> dispose of the e-mail.
>
> =====================
> 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

[Ergul, Emel A.]

Thanks very much Art!
emel


-----Original Message-----
From: Art Kendall [mailto:[hidden email]]
Sent: Wed 6/17/2009 5:58 PM
To: Ergul, Emel A.
Cc: [hidden email]
Subject: Re:

Since you do not have a large amount of data, I strongly suggest that
you enter it again.  1)Proofread your data dictionary 2)for as many
people you have to re-enter the data create new entry files using the
same dictionary.  3) have people re-enter the data  4)Use ADD files to
put the new set of files together into a second version.  5) Sort the
original entry file and the second entry file in the same order.


Then search the archives for COMPDS  for how to compare data sets.

Art Kendall
Social Research Consultants

Ergul, Emel A. wrote:
> Dear List:
>
> Again, my question might be seemingly simple even stupid for many of you who
> have been working on meta data. Please accept my apologies if it is so.
>
> CASE: I want to validate my data where the data entry were done with few other
> people. I have 900 cases and about 300 variables I have the hard copy of
values
> (correct) and values that had been entered (correct and incorrect). My plan is
> to take XX number of people randomly from the set and review the entries.
>
> QUESTION: How can I set up an upper limit for err for evidence of
in/validation?

> Do you know any good book or online resource on data validation I can read.
>
> I'm grateful as always for any help,
> E.
>
> PS: Majority of resources appeared online is for those who wants to set up a
> program prior to data entry for incorrect values. Mine are already there.
>
>
>
> The information in this e-mail is intended only for the person to whom it is
> addressed. If you believe this e-mail was sent to you in error and the e-mail
> contains patient information, please contact the Partners Compliance HelpLine

at
> http://www.partners.org/complianceline . If the e-mail was sent to you in
error
> but does not contain patient information, please contact the sender and
properly

> dispose of the e-mail.
>
> =====================
> 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