dichotomous data with repeated measures
Locked
 
|
Hi all,
I have a question concerning the analysis of a part of an experimental vignette study we did. We have asked children to evaluate and give justifications on some stories. Then we presented them with three counter probes and studied if they changed their opinion. I want to analyze the change of opinion for each story (4 in total) when peers, parents or reciprocity rules differ. These are coded as binary variables (change/no change) and measured within subjects. It looks likes this: child peers-story1 parents-story1 reciprocity-story1 etc 1 0 1 1 2 1 1 1 3 0 0 1 The main question I have is what probe (for each separate story) was most effective and whether probes differed in effectiveness. Basically, I just want to compare the proportions of each probe. In developmental psychological journals the counter-probe technique is, almost without exception, analyzed with repeated measures ANOVA with references showing ANOVA is robust with dichotomous data and is the preferred method over log-linear models (such as Lunney, 1970; Dágostino, 1971; Gaito, 1980). However, it still seems odd to me to use ANOVA and I feel there might be a better way, although I also see that log-linear models are problematic due to the repeated measure structure of my data. Does anybody have any suggestions? Thanks! |
Re: dichotomous data with repeated measures
|
|
Unless I'm missing something, it would seem to me that this would be a good candidate for an MLM. In SPSS this can be done via the mixed models generalized linear model with a binary logistic regression (i.e. log link).
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 J.S. Sent: Wednesday, September 26, 2012 8:43 AM To: [hidden email] Subject: dichotomous data with repeated measures Hi all, I have a question concerning the analysis of a part of an experimental vignette study we did. We have asked children to evaluate and give justifications on some stories. Then we presented them with three counter probes and studied if they changed their opinion. I want to analyze the change of opinion for each story (4 in total) when peers, parents or reciprocity rules differ. These are coded as binary variables (change/no change) and measured within subjects. It looks likes this: child peers-story1 parents-story1 reciprocity-story1 etc 1 0 1 1 2 1 1 1 3 0 0 1 The main question I have is what probe (for each separate story) was most effective and whether probes differed in effectiveness. Basically, I just want to compare the proportions of each probe. In developmental psychological journals the counter-probe technique is, almost without exception, analyzed with repeated measures ANOVA with references showing ANOVA is robust with dichotomous data and is the preferred method over log-linear models (such as Lunney, 1970; Dágostino, 1971; Gaito, 1980). However, it still seems odd to me to use ANOVA and I feel there might be a better way, although I also see that log-linear models are problematic due to the repeated measure structure of my data. Does anybody have any suggestions? Thanks! -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/dichotomous-data-with-repeated-measures-tp5715280.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: dichotomous data with repeated measures
|
|
In reply to this post by J.S.
Your description of the data structure is unclear. It sounds like each kid reads/hears multiple stories and gives a justification for each story and you then challenge the justification with three different counter probes. If true, then don't you have a three level dataset: probes within stories within kids? OR, a doubly repeated measures?
In terms of spss itself, I'd guess that GenLinMixed would be the best candidate procedure or an R routine running in spss. Outside of spss: Mplus but I'd guess sas or stata also would analyse the data. Gene Maguin -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J.S. Sent: Wednesday, September 26, 2012 9:43 AM To: [hidden email] Subject: dichotomous data with repeated measures Hi all, I have a question concerning the analysis of a part of an experimental vignette study we did. We have asked children to evaluate and give justifications on some stories. Then we presented them with three counter probes and studied if they changed their opinion. I want to analyze the change of opinion for each story (4 in total) when peers, parents or reciprocity rules differ. These are coded as binary variables (change/no change) and measured within subjects. It looks likes this: child peers-story1 parents-story1 reciprocity-story1 etc 1 0 1 1 2 1 1 1 3 0 0 1 The main question I have is what probe (for each separate story) was most effective and whether probes differed in effectiveness. Basically, I just want to compare the proportions of each probe. In developmental psychological journals the counter-probe technique is, almost without exception, analyzed with repeated measures ANOVA with references showing ANOVA is robust with dichotomous data and is the preferred method over log-linear models (such as Lunney, 1970; Dágostino, 1971; Gaito, 1980). However, it still seems odd to me to use ANOVA and I feel there might be a better way, although I also see that log-linear models are problematic due to the repeated measure structure of my data. Does anybody have any suggestions? Thanks! -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/dichotomous-data-with-repeated-measures-tp5715280.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 |
|
|
Thanks for your reply.
To clarify, indeed children were read to a story, asked for their evaluations/justification and then counter pobes were introduced one by one. I guess, a double repeated design.. I analyzed evaluations of the different stories with multivariate multilevel regression analyses. Now I am just interested as a validation of children's evaluations, how counterprobes influenced them. If I were to use anova, i would report planned contrast or post-hoc comparisons (e.g. "reciprocity is a more influential probe than peers in story 1.."). However, Anova does not seem appropriate (Do you agree?). I don't think I entirely understand how to use general linear mixed models here, does that not imply a continuous outcome variable? Or am I mistaken? (not used mixed models before) Thanks for your help in advance... |
Re: dichotomous data with repeated measures
|
|
No, a mixed model, HLM, or MLM (All different names for the same thing) can involve any of the general linear models. This includes logistic and probit links, i.e. logistic or probit regression. Using the Generalized Linear Models under Mixed models in SPSS allows you to run a logistic MLM with SPSS (Something not possible until recently in SPSS). It can also has been possible with SAS and HLM for some time now. I just worn that you watch your interpretation carefully. I've seen a fair bit of totally inaccurate interpretations of these models, even logistic regression models that have made it to press.
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 J.S. Sent: Wednesday, September 26, 2012 10:20 AM To: [hidden email] Subject: Re: dichotomous data with repeated measures Thanks for your reply. To clarify, indeed children were read to a story, asked for their evaluations/justification and then counter pobes were introduced one by one. I guess, a double repeated design.. I analyzed evaluations of the different stories with multivariate multilevel regression analyses. Now I am just interested as a validation of children's evaluations, how counterprobes influenced them. If I were to use anova, i would report planned contrast or post-hoc comparisons (e.g. "reciprocity is a more influential probe than peers in story 1.."). However, Anova does not seem appropriate (Do you agree?). I don't think I entirely understand how to use general linear mixed models here, does that not imply a continuous outcome variable? Or am I mistaken? (not used mixed models before) Thanks for your help in advance... -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/dichotomous-data-with-repeated-measures-tp5715280p5715287.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 |
Automatic reply: dichotomous data with repeated measures
|
|
I am on annual leave until the 29th of October and will respond to your email as soon as possible upon my return. Regards, Pawel |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |