CTables Test of Significance

Monday morning blank ...

I have a set of variables say different Jeep models like Compass, Wrangler, Cherokee .... and for each a respondent does a rating from Very Satisfied - Very Unsatisfied (Ordinal from 1 to 5) and want to do a test of sig between the rating for each model.

For the life of me am blanked. In Ctables can set that x'tab up but since I reposition the response codes from the brand to the column test of sig want work

What am I missing

Running V24.0.0.1

 

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

[Bruce Weaver]

To summarize:

1. You have an ordinal dependent variable (rating).
2. You have one categorical explanatory variable (Jeep Model, with levels Compass, Wrangler, etc.).
3. Each respondent rates each model, so you have repeated measures.

One way to address your question is via GENLIN, using a multinomial error distribution with a cumulative logit (CUMLOGIT) link function, and with generalized estimating equations (GEE) to account for the correlated nature of the repeated measures.  (Another way would be to estimate a multilevel model via GENLINMIXED, with Jeep Model at level 1 and Respondent at level 2.)

I am not a CTABLES expert, but when I look it up in the FM, here is what I find under the heading of Significance Testing:

"Custom Tables can perform the chi-square test of independence and pairwise comparisons of column
proportions for tables that contain at least one category variable in both the rows and the columns."

The Chi-square test of independence assumes one observation per respondent, and is not appropriate for your repeated measures data.

HTH.

WillBaileyz wrote

Monday morning blank ...

I have a set of variables say different Jeep models like Compass, Wrangler,
Cherokee .... and for each a respondent does a rating from Very Satisfied -
Very Unsatisfied (Ordinal from 1 to 5) and want to do a test of sig between
the rating for each model.

For the life of me am blanked. In Ctables can set that x'tab up but since I
reposition the response codes from the brand to the column test of sig want
work

What am I missing

Running V24.0.0.1

 


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

[Kornbrot, Diana]

think you can use cross tabs with Mantel Hanzel option with subjects

GEE is fine recommend you try probit as well as logit

GENLINMIXED only has multinomial option - not ordinal logit,

also i claims not to do mixed for repeated measu

as far as I can see, however you can run both probit and logit multinomial and see which give you lowest Aikake or Bayes goodness of fit. Bet its probit - but would be interested to know

Bruce if you know how to get ordinal or even multinomial logit in genlinmixed - please tell us, WITH SCRIPT. in v23 it said not -  multinomial not available

SPSS is DIRE fro ordinal it won’t even give you the means in either GEE or regression ordinal [plum] 

as a side line recession ordinal is preferable for between designs as you can fit scale as well as location

best

Diana

On 28 Mar 2017, at 02:59, Bruce Weaver <[hidden email]> wrote:

To summarize:

1. You have an ordinal dependent variable (rating).
2. You have one categorical explanatory variable (Jeep Model, with levels
Compass, Wrangler, etc.).
3. Each respondent rates each model, so you have repeated measures.

One way to address your question is via GENLIN, using a multinomial error
distribution with a cumulative logit (CUMLOGIT) link function, and with
generalized estimating equations (GEE) to account for the correlated nature
of the repeated measures.  (Another way would be to estimate a multilevel
model via GENLINMIXED, with Jeep Model at level 1 and Respondent at level
2.)

I am not a CTABLES expert, but when I look it up in the FM, here is what I
find under the heading of Significance Testing:

"Custom Tables can perform the chi-square test of independence and pairwise
comparisons of column
proportions for tables that contain at least one category variable in both
the rows and the columns."

The Chi-square test of independence assumes one observation per respondent,
and is not appropriate for your repeated measures data.

HTH.



WillBaileyz wrote

Monday morning blank ...

I have a set of variables say different Jeep models like Compass,
Wrangler,
Cherokee .... and for each a respondent does a rating from Very Satisfied
-
Very Unsatisfied (Ordinal from 1 to 5) and want to do a test of sig
between
the rating for each model. 

For the life of me am blanked. In Ctables can set that x'tab up but since
I
reposition the response codes from the brand to the column test of sig
want
work

What am I missing

Running V24.0.0.1




=====================
To manage your subscription to SPSSX-L, send a message to

LISTSERV@.UGA

(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

-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/CTables-Test-of-Significance-tp5734016p5734017.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

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

________________________________________
Professor Diana Kornbrot
Work
University of Hertfordshire
College Lane, Hatfield, Hertfordshire AL10 9AB, UK
[phone]
[hidden email]
http://dianakornbrot.wordpress.com/
 http://go.herts.ac.uk/Diana_Kornbrot
skype:  kornbrotme
Home
19 Elmhurst Avenue
London N2 0LT, UK
 [phone]                                                   

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

[Bruce Weaver]

Hi Diana.  I haven't actually tried that analysis with GENLINMIXED.  I was just supposing that surely, it must be able to do it.  

For GENLIN with GEE, here is an example of something I did in the past that the OP might be able to use as a template.

https://groups.google.com/forum/#!topic/comp.soft-sys.stat.spss/zkxR016mZxM

HTH.

Kornbrot, Diana wrote

think you can use cross tabs with Mantel Hanzel option with subjects

GEE is fine recommend you try probit as well as logit
GENLINMIXED only has multinomial option - not ordinal logit,
also i claims not to do mixed for repeated measu
as far as I can see, however you can run both probit and logit multinomial and see which give you lowest Aikake or Bayes goodness of fit. Bet its probit - but would be interested to know

Bruce if you know how to get ordinal or even multinomial logit in genlinmixed - please tell us, WITH SCRIPT. in v23 it said not -  multinomial not available

SPSS is DIRE fro ordinal it won’t even give you the means in either GEE or regression ordinal [plum]
as a side line recession ordinal is preferable for between designs as you can fit scale as well as location
best
Diana

On 28 Mar 2017, at 02:59, Bruce Weaver <[hidden email]<mailto:[hidden email]>> wrote:

To summarize:

1. You have an ordinal dependent variable (rating).
2. You have one categorical explanatory variable (Jeep Model, with levels
Compass, Wrangler, etc.).
3. Each respondent rates each model, so you have repeated measures.

One way to address your question is via GENLIN, using a multinomial error
distribution with a cumulative logit (CUMLOGIT) link function, and with
generalized estimating equations (GEE) to account for the correlated nature
of the repeated measures.  (Another way would be to estimate a multilevel
model via GENLINMIXED, with Jeep Model at level 1 and Respondent at level
2.)

I am not a CTABLES expert, but when I look it up in the FM, here is what I
find under the heading of Significance Testing:

"Custom Tables can perform the chi-square test of independence and pairwise
comparisons of column
proportions for tables that contain at least one category variable in both
the rows and the columns."

The Chi-square test of independence assumes one observation per respondent,
and is not appropriate for your repeated measures data.

HTH.



WillBaileyz wrote
Monday morning blank ...

I have a set of variables say different Jeep models like Compass,
Wrangler,
Cherokee .... and for each a respondent does a rating from Very Satisfied
-
Very Unsatisfied (Ordinal from 1 to 5) and want to do a test of sig
between
the rating for each model.

For the life of me am blanked. In Ctables can set that x'tab up but since
I
reposition the response codes from the brand to the column test of sig
want
work

What am I missing

Running V24.0.0.1




=====================
To manage your subscription to SPSSX-L, send a message to

LISTSERV@.UGA

(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





-----
--
Bruce Weaver
[hidden email]<mailto:[hidden email]>
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/CTables-Test-of-Significance-tp5734016p5734017.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com<http://nabble.com/>.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email]<mailto:[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

________________________________________
Professor Diana Kornbrot
Work
University of Hertfordshire
College Lane, Hatfield, Hertfordshire AL10 9AB, UK
[phone]
[hidden email]<mailto:[hidden email]>
http://dianakornbrot.wordpress.com/
 http://go.herts.ac.uk/Diana_Kornbrot
skype:  kornbrotme
Home
19 Elmhurst Avenue
London N2 0LT, UK
 [phone]




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

[Rich Ulrich]

Someone should mention that likert-type scorings are usually analyzed by assuming that

the scores are sufficiently equal-interval to justify ordinary ANOVA. By that token, you have

an ordinary repeated measures design.  Cautious testing between pairs will use paired t-tests.

Are you happy to describe the "average satisfaction" for a Jeep model?  If not, that would

justify considering the models with transformations.

To treat the rankings of each individual's scores -- a nonparametric, "ordinal" approach -- the

overall test would be Friedman's two-way ANOVA. Follow that with the non-parametric version

of a test.

Those logit or probit models will do better when you do not have large numbers of ties, that is,

when you start with continuous scores that always give unique values for ranks.

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of Willbaileyz @ E <[hidden email]>
Sent: Monday, March 27, 2017 3:52:22 PM
To: [hidden email]
Subject: CTables Test of Significance

 

Monday morning blank ...

I have a set of variables say different Jeep models like Compass, Wrangler, Cherokee .... and for each a respondent does a rating from Very Satisfied - Very Unsatisfied (Ordinal from 1 to 5) and want to do a test of sig between the rating for each model.

For the life of me am blanked. In Ctables can set that x'tab up but since I reposition the response codes from the brand to the column test of sig want work

What am I missing

Running V24.0.0.1

 

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

[Bruce Weaver]

Thom Baguley, whom some of you might know from other discussion forums, has written a blog post entitled "Beware the Friedman test!"

  https://seriousstats.wordpress.com/2012/02/14/friedman/

Here is an excerpt that captures one of Baguley's main problems with the Friedman test.

"The ranks for the Friedman test depend only on the order of scores within each participant – they completely ignore the differences between participants. This differs dramatically from the Wilcoxon test where information about the relative size of differences between participants is preserved. Zimmerman and Zumbo (1993) discuss this difference in procedures and explain that the Friedman test (devised by the noted economist and champion of the ‘free market’ Milton Friedman) is not really a form of ANOVA but an extension of the sign test. It is an impostor."

Cheers,
Bruce

Rich Ulrich wrote

Someone should mention that likert-type scorings are usually analyzed by assuming that

the scores are sufficiently equal-interval to justify ordinary ANOVA. By that token, you have

an ordinary repeated measures design.  Cautious testing between pairs will use paired t-tests.


Are you happy to describe the "average satisfaction" for a Jeep model?  If not, that would

justify considering the models with transformations.


To treat the rankings of each individual's scores -- a nonparametric, "ordinal" approach -- the

overall test would be Friedman's two-way ANOVA. Follow that with the non-parametric version

of a test.


Those logit or probit models will do better when you do not have large numbers of ties, that is,

when you start with continuous scores that always give unique values for ranks.

--

Rich Ulrich

________________________________
From: SPSSX(r) Discussion <[hidden email]> on behalf of Willbaileyz @ E <[hidden email]>
Sent: Monday, March 27, 2017 3:52:22 PM
To: [hidden email]
Subject: CTables Test of Significance


Monday morning blank ...

I have a set of variables say different Jeep models like Compass, Wrangler, Cherokee .... and for each a respondent does a rating from Very Satisfied - Very Unsatisfied (Ordinal from 1 to 5) and want to do a test of sig between the rating for each model.

For the life of me am blanked. In Ctables can set that x'tab up but since I reposition the response codes from the brand to the column test of sig want work

What am I missing

Running V24.0.0.1

===================== To manage your subscription to SPSSX-L, send a message to [hidden email]<mailto:[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