SPSSX Discussion

Chi-sq statistics

Classic

List

Threaded

8 messages Options

Wincy Chan

Chi-sq statistics

Dear all

I have a fundamental question about chi-square statistics in SPSS. For
observations n<5, theoretically we are not supposed to use Chi-Sq but
Fisher’s test. However, SPSS does generate a chi-sq value with significance
(with a warning message but no error message). I heard SPSS has adjusted the
chi-sq statistics – is that true? Can we report such a finding?

Many thanks

Wincy

　

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

John Amora

Re: Chi-sq statistics

Hi Wincy,

The following are stated in the SPSS help:

For a 2 × 2 table, Fisher's exact test is computed when a table that does not result from missing rows or columns in a larger table has a cell with an expected frequency of less than 5. Yates' corrected chi-square is computed for all other 2 × 2 tables.

For RxC table with small sample (especially table with large percentage of expected values below 5), you may use exact p-value. Exact p-values are available in SPSS Exact test module. Try this link: http://www.spss-sa.com/images/brochures/Exact_tests.pdf

Cheers,
Johnny

--- On Tue, 11/18/08, Wincy Chan <[hidden email]> wrote:

From: Wincy Chan <[hidden email]>
Subject: Chi-sq statistics
To: [hidden email]
Date: Tuesday, 18 November, 2008, 2:12 PM

Dear all

I have a fundamental question about chi-square statistics in SPSS. For
observations n<5, theoretically we are not supposed to use Chi-Sq but
Fisher’s test. However, SPSS does generate a chi-sq value with significance
(with a warning message but no error message). I heard SPSS has adjusted the
chi-sq statistics – is that true? Can we report such a finding?

Many thanks

Wincy

　

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

New Email addresses available on Yahoo!
Get the Email name you've always wanted on the new @ymail and @rocketmail.
Hurry before someone else does!
http://mail.promotions.yahoo.com/newdomains/ph/

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

Marta Garcia-Granero

Re: Chi-sq statistics

In reply to this post by Wincy Chan

Wincy Chan wrote:
> I have a fundamental question about chi-square statistics in SPSS. For
> observations n<5, theoretically we are not supposed to use Chi-Sq but
> Fisher’s test. However, SPSS does generate a chi-sq value with significance
> (with a warning message but no error message). I heard SPSS has adjusted the
> chi-sq statistics – is that true? Can we report such a finding?

SPSS computes Fisher's exact test in 2x2 contingency tables analysis.
For bigger contingency tables (RxC) or one sample chi-square test
(goodness of fit test) you need the EXACT TESTS module installed to get
the exact p-values. You don't mention the test you are using
(contingency or goodness of fit).

If you re not sure about the SPSS components you have installed, open a
blank syntax window , write:

SHOW LICENSE.

And run it.

HTH,
Marta García-Granero

--
For miscellaneous 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

Wincy Chan

Re: Chi-sq statistics

Thank you Marta, Johnny, and Gary for your reply. I obviously don't have the
Exact Test module and am on way to order it.
Cheers
Wincy

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Marta Garcia-Granero
Sent: Tuesday, November 18, 2008 4:51 PM
To: [hidden email]
Subject: Re: Chi-sq statistics

Wincy Chan wrote:
> I have a fundamental question about chi-square statistics in SPSS. For
> observations n<5, theoretically we are not supposed to use Chi-Sq but
> Fishers test. However, SPSS does generate a chi-sq value with
significance
> (with a warning message but no error message). I heard SPSS has adjusted
the
> chi-sq statistics is that true? Can we report such a finding?

SPSS computes Fisher's exact test in 2x2 contingency tables analysis.
For bigger contingency tables (RxC) or one sample chi-square test
(goodness of fit test) you need the EXACT TESTS module installed to get
the exact p-values. You don't mention the test you are using
(contingency or goodness of fit).

If you re not sure about the SPSS components you have installed, open a
blank syntax window , write:

SHOW LICENSE.

And run it.

HTH,
Marta García-Granero

--
For miscellaneous 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

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

Marta Garcia-Granero

Re: Chi-sq statistics

Wincy Chan wrote:
> Thank you Marta, Johnny, and Gary for your reply. I obviously don't have the
> Exact Test module and am on way to order it.
>
A couple of clarifications I forgot to mention in my previous message:

- Chi-square tests can be valid even when expected frequencies are <5,
provided that:
1) the lowest expected frequency is not <1
2) the percentage of cells with frequencies<5 is lower than 20%

- If you need an exact p-value for a 2xk contingency table (not bigger),
there is a freeware utility (not surprisingly, its name is
EXACT2XK.EXE), part of the PEPI package. It used to be downloaded from
www.sagebrushpress.com, but unfortunately it is not there anymore (the
domain is for sale, the page is full of worthless information now). I
think that the windows version of the package, WINPEPI, will do the
trick too. You can download WINPEPI from:
http://www.brixtonhealth.com/pepi4windows.html

HTH,
Marta García-Granero

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

Wincy Chan

Re: Chi-sq statistics

Dear Marta

Your information is very helpful. Thank you very much.

Regards
Wincy

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Marta Garcia-Granero
Sent: Wednesday, November 19, 2008 4:08 PM
To: [hidden email]
Subject: Re: Chi-sq statistics

Wincy Chan wrote:
> Thank you Marta, Johnny, and Gary for your reply. I obviously don't have
the
> Exact Test module and am on way to order it.
>
A couple of clarifications I forgot to mention in my previous message:

- Chi-square tests can be valid even when expected frequencies are <5,
provided that:
1) the lowest expected frequency is not <1
2) the percentage of cells with frequencies<5 is lower than 20%

- If you need an exact p-value for a 2xk contingency table (not bigger),
there is a freeware utility (not surprisingly, its name is
EXACT2XK.EXE), part of the PEPI package. It used to be downloaded from
www.sagebrushpress.com, but unfortunately it is not there anymore (the
domain is for sale, the page is full of worthless information now). I
think that the windows version of the package, WINPEPI, will do the
trick too. You can download WINPEPI from:
http://www.brixtonhealth.com/pepi4windows.html

HTH,
Marta García-Granero

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

Jims More

Re: Chi-sq statistics

In reply to this post by Marta Garcia-Granero

Hi Marta, Johnny, et al.,

I've tried to follow your suggestions to use Exact Test using my own data; 4 by 4 table, one is ordinal and the other is nominal, with 10 cells (62.5%) have expected count less than 5, and the minimum expected count is .12. SPSS 15 provides the exact p-values for likelihood ratio, Fisher's Exact Test, and Linear-by-Linear Association. Do I use Fisher's Exact Test?

Thanks.

Jims

--- On Wed, 19/11/08, Marta García-Granero <[hidden email]> wrote:

From: Marta García-Granero <[hidden email]>
Subject: Re: Chi-sq statistics
To: [hidden email]
Received: Wednesday, 19 November, 2008, 8:07 AM

Wincy Chan wrote:
> Thank you Marta, Johnny, and Gary for your reply. I obviously don't
have the
> Exact Test module and am on way to order it.
>
A couple of clarifications I forgot to mention in my previous message:

- Chi-square tests can be valid even when expected frequencies are <5,
provided that:
1) the lowest expected frequency is not <1
2) the percentage of cells with frequencies<5 is lower than 20%

- If you need an exact p-value for a 2xk contingency table (not bigger),
there is a freeware utility (not surprisingly, its name is
EXACT2XK.EXE), part of the PEPI package. It used to be downloaded from
www.sagebrushpress.com, but unfortunately it is not there anymore (the
domain is for sale, the page is full of worthless information now). I
think that the windows version of the package, WINPEPI, will do the
trick too. You can download WINPEPI from:
http://www.brixtonhealth.com/pepi4windows.html

HTH,
Marta García-Granero

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

Start your day with Yahoo!7 and win a Sony Bravia TV. Enter now http://au.docs.yahoo.com/homepageset/?p1=other&p2=au&p3=tagline

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

Marta Garcia-Granero

Re: Chi-sq statistics

Jims More wrote:

> Hi Marta, Johnny, et al.,
>
> I've tried to follow your suggestions to use Exact Test using my own
> data; 4 by 4 table, one is ordinal and the other is nominal, with 10
> cells (62.5%) have expected count less than 5, and the minimum
> expected count is .12. SPSS 15 provides the exact p-values for
> likelihood ratio, Fisher's Exact Test, and Linear-by-Linear
> Association. Do I use Fisher's Exact Test?
>
>
> <http://au.rd.yahoo.com/hppromo/mail/tagline2/*http://au.docs.yahoo.com/homepageset/?p1=other&p2=au&p3=tagline%20>
>
>

Use Fisher's (complete name, BTW, is fisher-Freeman-Halton) test.

Regards,
Marta

--
For miscellaneous 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