estimating reliability when there is a set of items in a scale rated by an incomplete set of raters
Locked
estimating reliability when there is a set of items in a scale rated by an incomplete set of raters
 
|
Does anyone know of a link that I could pass on that shows how to
estimate reliability when there are a set of summative items and a set of raters but for some subjects a rater is missing for all of the items. This is a doubly repeated design -- Art Kendall Social Research Consultants ===================== 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
Social Research Consultants |
Re: estimating reliability when there is a set of items in a scale rated by an incomplete set of raters
|
Are you looking for something that addresses the issue of inter-rater reliability with missing data? In other words, the rater is missing within a domain. If so, I'd have to relook at some papers I have on the topic, but my knee jerk reaction is I'm not sure I've read too much that specifically addresses missing rater information and inter-rater reliability.
http://www.afhayes.com/public/cmm2007.pdf Let me know if this paper helps. It discusses the K-Alpha approach, and does mention its general ability to be used with missing data. I'm just not totally clear if this is what you are looking for. This person may also have to adapt concepts from a set of area's to directly address the problem they have. For instance, we know that certain approaches are more robust to missing data under specific conditions, so applying this within the domain they are dealing with. Matthew J Poes Research Data Specialist Center for Prevention Research and Development University of Illinois 510 Devonshire Dr. Champaign, IL 61820 Phone: 217-265-4576 email: [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Art Kendall Sent: Wednesday, July 11, 2012 11:14 AM To: [hidden email] Subject: estimating reliability when there is a set of items in a scale rated by an incomplete set of raters Does anyone know of a link that I could pass on that shows how to estimate reliability when there are a set of summative items and a set of raters but for some subjects a rater is missing for all of the items. This is a doubly repeated design -- Art Kendall Social Research Consultants ===================== 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 |
|
|
In reply to this post by Art Kendall
The mixed procedure. I think I've posted a message about using the mixed procedure to estimate ICC.
Ryan On Jul 11, 2012, at 12:13 PM, Art Kendall <[hidden email]> wrote: > Does anyone know of a link that I could pass on that shows how to > estimate reliability when > there are a set of summative items and a set of raters but for some > subjects a rater is missing for all of the items. This is a doubly > repeated design > > -- > Art Kendall > Social Research Consultants > > ===================== > 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 |
Re: estimating reliability when there is a set of items in a scale rated by an incomplete set of raters
|
|
I thought I had seen
something on this list but did not recall enough to go through
the archives. I should have better luck now that I have a
name. "ICC" and "reliability" was not specific enough.
When a rater rated a subject (s)he did all of the 7 rating items, otherwise the 7 variables have missing data. So there are subjects (random) * rating item (random) * rater (random possibly could be considered fixed). So there are components of scale (inter-item) reliability and of inter-rater reliability. Art Kendall Social Research ConsultantsOn 7/11/2012 12:30 PM, Ryan Black wrote: The mixed procedure. I think I've posted a message about using the mixed procedure to estimate ICC. Ryan On Jul 11, 2012, at 12:13 PM, Art Kendall [hidden email] wrote:Does anyone know of a link that I could pass on that shows how to estimate reliability when there are a set of summative items and a set of raters but for some subjects a rater is missing for all of the items. This is a doubly repeated design -- Art Kendall Social Research Consultants ===================== 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 ===================== 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
Social Research Consultants |
|
|
A linear mixed model can certainly be used to estimate an ICC even if every rater did not rate each subject. Is that what you're asking. If so, without much more information, FWIW that would be my general reaction. -Ryan
On Wed, Jul 11, 2012 at 6:43 PM, Art Kendall <[hidden email]> wrote:
|
Re: estimating reliability when there is a set of items in a scale rated by an incomplete set of raters
|
|
In reply to this post by Art Kendall
It finally occurred to me that the topic was included in my old Stats-FAQ,
which had collected good posts from 1995 to about 1999. It is no longer on the web, but I still carry the pieces between my computers. This is not the exact model, but the principle of adjusting k should be the same. I haven't check the 1958 reference. (Reformated here as text, plust the change of one symbol. Google Groups does not find this original post for me, even by Message-ID. In my efforts.) =====================Michael Bailey, 11 Oct 1995========ssc From: [hidden email] (Michael Bailey) Subject: Re: reliability query Message-ID: <[hidden email]> In article <[hidden email]>, Jeffrey Berman <[hidden email]> wrote: <snip> > The F ratio from this ANOVA would be the test of the intraclass correlation. > The only complication in calculating the intraclass correlation itself is > that you appear to have unequal numbers of estimates within each company. A > method for computing the intraclass correlation in the case of unequal class > membership is outlined in the classic reference by Haggard (1958, chap. 2). > I would be interested if others on the list have more recent references for > calculating the intraclass correlation in the case of unequal class membership. Thanks Jeff, you're right, and the correct way to do it is on page 14 of Haggard. Let, R=intraclass correlation, BCMS=between classes mean square, WMS=mean square within, c=number of classes, and ki=number in the ith group. THen: R = BCMS - WMS __________ BCMS + (k_a -1)WMS where k_a = ( 1 /(c-1) )* (sigma ki - (sigma (ki**2))/sigma ki) (sigma means summation) ==========end of cite from Bailey's post. -- Rich Ulrich > Date: Wed, 11 Jul 2012 12:13:55 -0400 > From: [hidden email] > Subject: estimating reliability when there is a set of items in a scale rated by an incomplete set of raters > To: [hidden email] > > Does anyone know of a link that I could pass on that shows how to > estimate reliability when > there are a set of summative items and a set of raters but for some > subjects a rater is missing for all of the items. This is a doubly > repeated design > > -- > Art Kendall > Social Research Consultants |
Re: estimating reliability when there is a set of items in a scale rated by an incomplete set of raters
|
|
In reply to this post by Ryan
A linear mixed model can certainly be used to estimate an ICC even if every rater did not rate each subject.That was my understanding. I was looking for a link I can pass on that explains how to to that in SPSS. Especially when there are two repeated factors, items and raters, i.e., both internal consistency and agreement. Now that I found your name I was able to find your post for 1 rating variable which I will certainly pass on. Art Kendall Social Research ConsultantsOn 7/11/2012 10:05 PM, R B wrote: A linear mixed model can certainly be used to estimate an ICC even if every rater did not rate each subject. Is that what you're asking. If so, without much more information, FWIW that would be my general reaction. -Ryan ===================== 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
Social Research Consultants |
Specifying higher-order terms for Logistic Regression
|
|
In reply to this post by Rich Ulrich
Is there any way to specify higher-order terms for logistic regression without calculating new variables? This syntax, allowing an interaction term to be specified, works fine: LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 /METHOD= ENTER var1 var2 var1*var2 . But this syntax, which is logically the same, gives the error message “Variable named is already in the BY interaction”; LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 /METHOD= ENTER var1 var2 var1*var1 . I’ve tried various alternatives such as including a “var1 BY var1” term on the variables subcommand, but I get the same error message. It seems illogical as other procedures like MIXED happily accept higher order terms like var1*var1 as well as interactions like var1*var2 in the list of independent variables. Garry Gelade PS. When is SPSS going to allow interaction terms/higher order terms to be specified directly in linear regression? |
Re: Specifying higher-order terms for Logistic Regression
Administrator
|
Hi Garry. Don't forget that in the REGRESSION procedure, you can't even use "var1*var2" -- you have to compute a new variable that is equal to the product. In other words, functionality that is available in some of the newer procedures just ain't there in many of the older ones.
Since your model is an ordinary binary logistic regression, you could try running it via NOMREG or GENLIN (with a logit link & binomial error distribution). I don't remember off the top of my head if NOMREG supports "var1*var1", but I'm fairly sure GENLIN does. HTH.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
Re: Specifying higher-order terms for Logistic Regression
|
|
And remember that the SPSSINC CREATE DUMMIES
extension command will compute and label up to three-way interaction variables
for categorical and scale variables for you and define macros to make using
them easy.
Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] new phone: 720-342-5621 From: Bruce Weaver <[hidden email]> To: [hidden email] Date: 07/13/2012 07:08 AM Subject: Re: [SPSSX-L] Specifying higher-order terms for Logistic Regression Sent by: "SPSSX(r) Discussion" <[hidden email]> Hi Garry. Don't forget that in the REGRESSION procedure, you can't even use "var1*var2" -- you have to compute a new variable that is equal to the product. In other words, functionality that is available in some of the newer procedures just ain't there in many of the older ones. Since your model is an ordinary binary logistic regression, you could try running it via NOMREG or GENLIN (with a logit link & binomial error distribution). I don't remember off the top of my head if NOMREG supports "var1*var1", but I'm fairly sure GENLIN does. HTH. Garry Gelade wrote > > Is there any way to specify higher-order terms for logistic regression > without calculating new variables? > > > > This syntax, allowing an interaction term to be specified, works fine: > > > > LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 > > /METHOD= ENTER var1 var2 var1*var2 . > > > > But this syntax, which is logically the same, gives the error message > "Variable named is already in the BY interaction"; > > LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 > > /METHOD= ENTER var1 var2 var1*var1 . > > > > I've tried various alternatives such as including a "var1 BY var1" term on > the variables subcommand, but I get the same error message. > > > > It seems illogical as other procedures like MIXED happily accept higher > order terms like var1*var1 as well as interactions like var1*var2 in the > list of independent variables. > > > > > > Garry Gelade > > > > PS. When is SPSS going to allow interaction terms/higher order terms to be > specified directly in linear regression? > ----- -- 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/estimating-reliability-when-there-is-a-set-of-items-in-a-scale-rated-by-an-incomplete-set-of-raters-tp5714139p5714192.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 |
Re: Specifying higher-order terms for Logistic Regression
|
|
In reply to this post by Garry Gelade
Garry, The documentation shows that this should work. LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 var1 by var2 /METHOD= ENTER var1 var2 var1 by var1 . Beyond that I agree that it’s really odd how the within-command subcommands differ so much from procedure to procedure. Mixed, glm, unianova are the most consistent; regression, logistic regression, plum and nomreg and maybe some of the newer are the least consistent. Thought was put into these decisions. While I don’t understand the underlying thinking; I wonder if it does’t reflect historical differences. Gene Maguin From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Garry Gelade Is there any way to specify higher-order terms for logistic regression without calculating new variables? This syntax, allowing an interaction term to be specified, works fine: LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 /METHOD= ENTER var1 var2 var1*var2 . But this syntax, which is logically the same, gives the error message “Variable named is already in the BY interaction”; LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 /METHOD= ENTER var1 var2 var1*var1 . I’ve tried various alternatives such as including a “var1 BY var1” term on the variables subcommand, but I get the same error message. It seems illogical as other procedures like MIXED happily accept higher order terms like var1*var1 as well as interactions like var1*var2 in the list of independent variables. Garry Gelade PS. When is SPSS going to allow interaction terms/higher order terms to be specified directly in linear regression? |
Re: Specifying higher-order terms for Logistic Regression
|
|
In reply to this post by Garry Gelade
Thanks, but this only seems to work if you want an interaction betyween two
different variables (var1*var2). But my problem is I want to specify the square of a variable (var1*var1). Garry -----Original Message----- From: M. Kruger [mailto:[hidden email]] Sent: 13 July 2012 14:24 To: Garry Gelade Cc: [hidden email] Subject: Re: Specifying higher-order terms for Logistic Regression You can use the drop down menus to specify higher order interactions. In the LR window when you specify covariates, there is a button '> a*b>'. Just hold the CTRL key as you select the variables that you want to be used in the interaction. I have used this button for years. No syntax required... MK ===================== 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 |
Re: Specifying higher-order terms for Logistic Regression
|
|
In reply to this post by Maguin, Eugene
Hi Eugene I presume you mean: LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 var1 by var1 /METHOD= ENTER var1 var2 var1 by var1 . But its not working on my system (SPSS 18, 64-bit by the way). Nor is: LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 var1 by var1 /METHOD= ENTER var1 var2 var1* var1 . I’m sure you’re right about the history – it looks like SPSS has never revisited some of the older procedures. Not so bad for us who have been using SPSS for a long time, but it must be difficult for newbies to grasp the different rules for specifying dependent variables. It would be good if SPSS could standardize that bit of the language across all procedures. Garry From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Maguin, Eugene Garry, The documentation shows that this should work. LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 var1 by var2 /METHOD= ENTER var1 var2 var1 by var1 . Beyond that I agree that it’s really odd how the within-command subcommands differ so much from procedure to procedure. Mixed, glm, unianova are the most consistent; regression, logistic regression, plum and nomreg and maybe some of the newer are the least consistent. Thought was put into these decisions. While I don’t understand the underlying thinking; I wonder if it does’t reflect historical differences. Gene Maguin From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Garry Gelade Is there any way to specify higher-order terms for logistic regression without calculating new variables? This syntax, allowing an interaction term to be specified, works fine: LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 /METHOD= ENTER var1 var2 var1*var2 . But this syntax, which is logically the same, gives the error message “Variable named is already in the BY interaction”; LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 /METHOD= ENTER var1 var2 var1*var1 . I’ve tried various alternatives such as including a “var1 BY var1” term on the variables subcommand, but I get the same error message. It seems illogical as other procedures like MIXED happily accept higher order terms like var1*var1 as well as interactions like var1*var2 in the list of independent variables. Garry Gelade PS. When is SPSS going to allow interaction terms/higher order terms to be specified directly in linear regression? |
Re: Specifying higher-order terms for Logistic Regression
|
|
In reply to this post by Bruce Weaver
Hi Bruce
Good idea. NOMREG does support var1*var1 by the way. Thanks Garry -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: 13 July 2012 13:59 To: [hidden email] Subject: Re: Specifying higher-order terms for Logistic Regression Hi Garry. Don't forget that in the REGRESSION procedure, you can't even use "var1*var2" -- you have to compute a new variable that is equal to the product. In other words, functionality that is available in some of the newer procedures just ain't there in many of the older ones. Since your model is an ordinary binary logistic regression, you could try running it via NOMREG or GENLIN (with a logit link & binomial error distribution). I don't remember off the top of my head if NOMREG supports "var1*var1", but I'm fairly sure GENLIN does. HTH. Garry Gelade wrote > > Is there any way to specify higher-order terms for logistic regression > without calculating new variables? > > > > This syntax, allowing an interaction term to be specified, works fine: > > > > LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 > > /METHOD= ENTER var1 var2 var1*var2 . > > > > But this syntax, which is logically the same, gives the error message > "Variable named is already in the BY interaction"; > > LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 > > /METHOD= ENTER var1 var2 var1*var1 . > > > > I've tried various alternatives such as including a "var1 BY var1" > term on the variables subcommand, but I get the same error message. > > > > It seems illogical as other procedures like MIXED happily accept > higher order terms like var1*var1 as well as interactions like > var1*var2 in the list of independent variables. > > > > > > Garry Gelade > > > > PS. When is SPSS going to allow interaction terms/higher order terms > to be specified directly in linear regression? > ----- -- 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/estimating-reliability-when-th ere-is-a-set-of-items-in-a-scale-rated-by-an-incomplete-set-of-raters-tp5714 139p5714192.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 ===================== 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 |
Re: Specifying higher-order terms for Logistic Regression
|
|
In reply to this post by Jon K Peck
But I don’t think it can create an interaction of a scale variable with itself (i.e var1*var1) can it? Garry From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Jon K Peck And remember that the SPSSINC CREATE DUMMIES extension command will compute and label up to three-way interaction variables for categorical and scale variables for you and define macros to make using them easy.
|
Re: Specifying higher-order terms for Logistic Regression
|
|
It's true that this can't be specified
in the dialog box, but you can do it in syntax.
SPSSINC CREATE DUMMIES VARIABLE=salary salary ROOTNAME2=twoway /OPTIONS ORDER=A USEVALUELABELS=YES USEML=YES OMITFIRST=YES MACRONAME2="!twoway". However, the command isn't currently smart enough to omit the redundant terms. If variable = salary salary prevexp, you get a redundant salary * prevexp interaction variable, but you can just ignore that term. I take that as a bug as also the dialog box issue. Regards, Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] new phone: 720-342-5621 From: Garry Gelade <[hidden email]> To: [hidden email] Date: 07/13/2012 08:26 AM Subject: Re: [SPSSX-L] Specifying higher-order terms for Logistic Regression Sent by: "SPSSX(r) Discussion" <[hidden email]> But I don’t think it can create an interaction of a scale variable with itself (i.e var1*var1) can it? Garry From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Jon K Peck Sent: 13 July 2012 14:11 To: [hidden email] Subject: Re: Specifying higher-order terms for Logistic Regression And remember that the SPSSINC CREATE DUMMIES extension command will compute and label up to three-way interaction variables for categorical and scale variables for you and define macros to make using them easy. Jon Peck (no "h") aka Kim Senior Software Engineer, IBM peck@... new phone: 720-342-5621 From: Bruce Weaver <bruce.weaver@...> To: [hidden email] Date: 07/13/2012 07:08 AM Subject: Re: [SPSSX-L] Specifying higher-order terms for Logistic Regression Sent by: "SPSSX(r) Discussion" <[hidden email]> Hi Garry. Don't forget that in the REGRESSION procedure, you can't even use "var1*var2" -- you have to compute a new variable that is equal to the product. In other words, functionality that is available in some of the newer procedures just ain't there in many of the older ones. Since your model is an ordinary binary logistic regression, you could try running it via NOMREG or GENLIN (with a logit link & binomial error distribution). I don't remember off the top of my head if NOMREG supports "var1*var1", but I'm fairly sure GENLIN does. HTH. Garry Gelade wrote > > Is there any way to specify higher-order terms for logistic regression > without calculating new variables? > > > > This syntax, allowing an interaction term to be specified, works fine: > > > > LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 > > /METHOD= ENTER var1 var2 var1*var2 . > > > > But this syntax, which is logically the same, gives the error message > "Variable named is already in the BY interaction"; > > LOGISTIC REGRESSION VARIABLES goal WITH var1 var2 > > /METHOD= ENTER var1 var2 var1*var1 . > > > > I've tried various alternatives such as including a "var1 BY var1" term on > the variables subcommand, but I get the same error message. > > > > It seems illogical as other procedures like MIXED happily accept higher > order terms like var1*var1 as well as interactions like var1*var2 in the > list of independent variables. > > > > > > Garry Gelade > > > > PS. When is SPSS going to allow interaction terms/higher order terms to be > specified directly in linear regression? > ----- -- Bruce Weaver bweaver@... 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/estimating-reliability-when-there-is-a-set-of-items-in-a-scale-rated-by-an-incomplete-set-of-raters-tp5714139p5714192.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSX-L, send a message to LISTSERV@... (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 |
Re: Specifying higher-order terms for Logistic Regression
|
Administrator
|
In reply to this post by Garry Gelade
I've just verified that NOMREG can handle var1*var1. E.g., the following runs with no problems:
GET FILE='C:\SPSSdata\1991 U.S. General Social Survey.sav'. NOMREG sex (BASE=LAST ORDER=ASCENDING) WITH age educ /MODEL=age educ age*age educ*educ age*educ /STEPWISE=PIN(.05) POUT(0.1) MINEFFECT(0) RULE(SINGLE) ENTRYMETHOD(LR) REMOVALMETHOD(LR) /INTERCEPT=INCLUDE /PRINT=PARAMETER SUMMARY LRT CPS STEP MFI. And of course, when the outcome variable is dichotomous, NOMREG produces the same model as LOGISTIC REGRESSION.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
Automatic reply: Specifying higher-order terms for Logistic Regression
|
|
I will be out of the officethis afternoon. If you have questions about the ADE system, please call Craig at 1-800-334-1918. Any other questions,
please contact Dan Hall, at 331-0415. Thanks! |
Re: Specifying higher-order terms for Logistic Regression
|
Administrator
|
In reply to this post by Garry Gelade
Oh, sorry. I posted my last message before spotting this one of yours!
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
| Free forum by Nabble | Edit this page |