Agreement (homogeneity) between subsample
Locked
Agreement (homogeneity) between subsample
 
|
Dear all,
I tested my SEModel in a full sample and in two subsamples (in the same country, same culture). My question is if a need to show before estimating my model that there is agreement (homogeneity I think) between the subsamples and how if yes. NB : the respondents are different from one subsample to another. Best regards, Mbaye |
Re: Agreement (homogeneity) between subsample
Administrator
|
What precisely do you mean by Homogeneity? This could mean MANY things!!!
Equal means? Equal Covariance matrices? (????) This may *NOT* be the best place to get answers about SEM. SEMNET might be a more likely resource: http://bama.ua.edu/archives/semnet.html
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
Re: Agreement (homogeneity) between subsample
|
|
In reply to this post by Mbaye Fall Diallo
Dear Mbaye Not sure what you mean by homogeneity, but I’m guessing you mean whether model parameters vary between groups. This is straightforward in principle. You estimate two multigroup models. In one model, you allow the parameters (eg measurement weights) to differ between the two samples. You then constrain the parameters to be equal across the groups. If the constrained model has a significantly worse fit than the unconstrained model, the parameters are significantly different in the two samples. You might want to consult a textbook on SEM or the user guide for your SEM software for more information. Regards Garry Gelade From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mbaye Fall Diallo Dear all, I tested my SEModel in a full sample and in two subsamples (in the same country, same culture). My question is if a need to show before estimating my model that there is agreement (homogeneity I think) between the subsamples and how if yes. NB : the respondents are different from one subsample to another. |
Re: Agreement (homogeneity) between subsample
|
|
In reply to this post by Mbaye Fall Diallo
There are two possibilities. One is to test for differences between the covariance matrices, which can be done with most SEM procedures. The second is to show invariance of the model across subsamples. This is a multiple group type structural model. What software are you using? Dr. Paul R. Swank, Children's Learning Institute Professor, Department of Pediatrics, Medical School Adjunct Professor, School of Public Health University of Texas Health Science Center-Houston From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mbaye Fall Diallo Dear all, I tested my SEModel in a full sample and in two subsamples (in the same country, same culture). My question is if a need to show before estimating my model that there is agreement (homogeneity I think) between the subsamples and how if yes. NB : the respondents are different from one subsample to another. |
Re: Agreement (homogeneity) between subsample
|
|
Dear Paul,
One of my doctoral thesis referees asked this question in his report. As I will defend my thesis in less than two weeks I wanted to know more about this. In my thesis I used Levene test of variance homogeneity. I also tested for configural and metric invariance in a multiple group analysis. I am using AMOS. Thank you for the feedback, Mbaye, Date: Mon, 21 Nov 2011 10:42:07 -0600 From: [hidden email] Subject: Re: Agreement (homogeneity) between subsample To: [hidden email] There are two possibilities. One is to test for differences between the covariance matrices, which can be done with most SEM procedures. The second is to show invariance of the model across subsamples. This is a multiple group type structural model. What software are you using?
Dr. Paul R. Swank, Children's Learning Institute Professor, Department of Pediatrics, Medical School Adjunct Professor, School of Public Health University of Texas Health Science Center-Houston
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mbaye Fall Diallo
Dear all, I tested my SEModel in a full sample and in two subsamples (in the same country, same culture). My question is if a need to show before estimating my model that there is agreement (homogeneity I think) between the subsamples and how if yes. NB : the respondents are different from one subsample to another.
|
Re: Agreement (homogeneity) between subsample
|
|
In reply to this post by Garry Gelade
Hi Garry,
That is the problem. I don't know what he meant exactly by homogeneity. May be I did not see that I tested for invariance. So I will prepare a table with details about invariance testing. Besides, as my sub samples are from the same culture, the same country, I am not sure that invariance testing was a great issue. Thank you. Best regards, Mbaye Fall Diallo. CERGAM - Management International Université Paul Cézanne/Aix-Marseille 3. Pensez à l'environnement avant d'imprimer cette page. Un petit geste peut avoir de grands effets. From: [hidden email] To: [hidden email] Subject: RE: Agreement (homogeneity) between subsample Date: Mon, 21 Nov 2011 16:54:09 +0000 Well if you’ve tested for invariance, I don’t know what else he wants. Regards Garry From: Mbaye Fall Diallo [mailto:[hidden email]]
Dear Garry, One of my doctoral thesis referee asked this question in his report. As will defend my thesis in less than two weeks I wanted to kwno more about this. In my thesis I used Levene test of variance homogeneity. I also tested for invariance (configural and metric). So I don't understand very well what he meant by homogeneity. Thank you for the clarifications. Best,
Date: Mon, 21 Nov 2011 12:02:14 +0000 Dear Mbaye
Not sure what you mean by homogeneity, but I’m guessing you mean whether model parameters vary between groups. This is straightforward in principle. You estimate two multigroup models. In one model, you allow the parameters (eg measurement weights) to differ between the two samples. You then constrain the parameters to be equal across the groups. If the constrained model has a significantly worse fit than the unconstrained model, the parameters are significantly different in the two samples.
You might want to consult a textbook on SEM or the user guide for your SEM software for more information.
Regards Garry Gelade
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mbaye Fall Diallo
Dear all, I tested my SEModel in a full sample and in two subsamples (in the same country, same culture). My question is if a need to show before estimating my model that there is agreement (homogeneity I think) between the subsamples and how if yes. NB : the respondents are different from one subsample to another.
|
|
|
In reply to this post by Mbaye Fall Diallo
Get this book.
Ryan
On Mon, Nov 21, 2011 at 12:15 PM, Mbaye Fall Diallo <[hidden email]> wrote:
|
Re: Agreement (homogeneity) between subsample
|
|
Hi Ryan,
I already have this book. Thank you. Best Mbaye, Date: Mon, 21 Nov 2011 12:45:26 -0500 From: [hidden email] Subject: Re: Agreement (homogeneity) between subsample To: [hidden email] Get this book. Ryan On Mon, Nov 21, 2011 at 12:15 PM, Mbaye Fall Diallo <[hidden email]> wrote:
|
Re: Agreement (homogeneity) between subsample
|
|
In reply to this post by Mbaye Fall Diallo
Hi Paul,
Thank you. I think I will be able to defend my position as I performed invariance test successfully. Best regards, Mbaye Fall Diallo. CERGAM - Management International Université Paul Cézanne/Aix-Marseille 3. Pensez à l'environnement avant d'imprimer cette page. Un petit geste peut avoir de grands effets. From: [hidden email] To: [hidden email] Date: Mon, 21 Nov 2011 11:47:25 -0600 Subject: RE: Agreement (homogeneity) between subsample I know how to test for differences between the variance covariance matrices in LISREL and Mplus but I have never used AMOS. However, I would be surprised if it couldn’t be done. The SEM list serve have a number of members who use AMOS and should be a good resource for you. Of course, if you have invariance in the model across groups, you should be okay.
Paul
Dr. Paul R. Swank, Children's Learning Institute Professor, Department of Pediatrics, Medical School Adjunct Professor, School of Public Health University of Texas Health Science Center-Houston
From: Mbaye Fall Diallo [mailto:[hidden email]]
Dear Paul, One of my doctoral thesis referees asked this question in his report. As I will defend my thesis in less than two weeks I wanted to know more about this. In my thesis I used Levene test of variance homogeneity. I also tested for configural and metric invariance in a multiple group analysis. I am using AMOS. Thank you for the feedback, Date: Mon, 21 Nov 2011 10:42:07 -0600 There are two possibilities. One is to test for differences between the covariance matrices, which can be done with most SEM procedures. The second is to show invariance of the model across subsamples. This is a multiple group type structural model. What software are you using?
Dr. Paul R. Swank, Children's Learning Institute Professor, Department of Pediatrics, Medical School Adjunct Professor, School of Public Health University of Texas Health Science Center-Houston
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mbaye Fall Diallo
Dear all, I tested my SEModel in a full sample and in two subsamples (in the same country, same culture). My question is if a need to show before estimating my model that there is agreement (homogeneity I think) between the subsamples and how if yes. NB : the respondents are different from one subsample to another.
|
| Free forum by Nabble | Edit this page |