[R] Mancova with R

John Fox jfox at mcmaster.ca
Wed Apr 17 17:45:50 CEST 2013


Dear Remi,

On Wed, 17 Apr 2013 08:23:07 -0700 (PDT)
 Rémi Lesmerises <remilesmerises at yahoo.ca> wrote:
> Dear John,
> 
> Thanks for your comments! But when I tried your suggestion, the output was as the following:
> 
> 
>  Response Dist_arbre :
>             Df     Sum Sq    Mean Sq F value    Pr(>F)    
> Poids        1 0.00010398 0.00010398  6.2910 0.0364733 *  
> Age          1 0.00005202 0.00005202  3.1476 0.1139652    
> ---
> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 
> 
> I have the P-value but not the direction of the relationship, information that I had with  "lm(as.matrix(Y) ~  x+z)". I could combine the results of these two tests, but it seems inelegant to me. Moreover I didn't have a total significance test as with a true MANCOVA. 
> 
> An idea?!

Yes, try what I suggested. The output that you show here isn't from the the Anova() function in the car package. As well, you might find it useful to read the on-line appendix on multivariate linear models, at <http://socserv.socsci.mcmaster.ca/jfox/Books/Companion/appendix/Appendix-Multivariate-Linear-Models.pdf>, from the book with which the car package is associated.

Best,
 John

> 
> Rémi Lesmerises, biol. M.Sc.,
> Candidat Ph.D. en Biologie
> Université du Québec à Rimouski
> 300, allée des Ursulines
> remilesmerises at yahoo.ca
> 
> 
> 
> ________________________________
>  De : John Fox <jfox at mcmaster.ca>
> À : Rémi Lesmerises <remilesmerises at yahoo.ca> 
> Cc : "r-help at r-project.org" <r-help at r-project.org> 
> Envoyé le : mercredi 17 avril 2013 10h54
> Objet : Re: [R] Mancova with R
>  
> 
> Dear Remi,
> 
> Take a look at the Anova() function in the car package. In your case, you could use
> 
> Anova(lm(as.matrix(Y) ~  x + z))
> 
> or, for more detail,
> 
> summary(Anova(lm(as.matrix(Y) ~  x + z)))
> 
> I hope this helps,
> John
> 
> ------------------------------------------------
> John Fox
> Sen. William McMaster Prof. of Social Statistics
> Department of Sociology
> McMaster University
> Hamilton, Ontario, Canada
> http://socserv.mcmaster.ca/jfox/
>     
> On Wed, 17 Apr 2013 07:47:27 -0700 (PDT)
> Rémi Lesmerises <remilesmerises at yahoo.ca> wrote:
> > Dear all,
> > 
> > I'm trying to compare two sets of variables, the first set is composed exclusively of numerical variables and the second regroups factors and numerical variables. I can't use a Manova because of this inclusion of numerical variables in the second set. The solution should be to perform a Mancova, but I didn't find any package that allow this type of test.
> > 
> > I've already looked in this forum and on the net to find answers, but the only thing I've found is the following:
> > 
> > 
> > lm(as.matrix(Y) ~  x+z)
> > x and z could be numerical and factors. The problem with that is it actually only perform a succession of lm (or glm), one for each numerical variable contained in the Y matrix. It is not a true MANCOVA that do a significance test (most often a Wald test) for the overall two sets comparison. Such a test is available in SPSS and SAS, but I really want to stay in R! Someone have any idea?
> > 
> > Thanks in advance for your help!
> >  
> > Rémi Lesmerises, biol. M.Sc.,
> > Candidat Ph.D. en Biologie
> > Université du Québec à Rimouski
> > remilesmerises at yahoo.ca
> > 
> >     [[alternative HTML version deleted]]
> > 

------------------------------------------------
John Fox
Sen. William McMaster Prof. of Social Statistics
Department of Sociology
McMaster University
Hamilton, Ontario, Canada
http://socserv.mcmaster.ca/jfox/



More information about the R-help mailing list