Hello! I?m fitting a model with glm(family binomial). The best model counts 9 Variables and includes an interaction term that was generated by the product of to continuous variables (a*b). All variables are correlated under a value of 0.7 (Spearman rank order) While the estimates of both main effects are negativ, the resulting interaction term is positiv. This change of sign makes it difficult to interpret the model and above all, is this perhaps due to a bad variable choice ? Thanks a lot for helping Christian _______________________________________________________________________ Viren-Scan f?r Ihren PC! Jetzt f?r jeden. Sofort, online und kostenlos. Gleich testen! http://www.pc-sicherheit.web.de/freescan/?mc=022222
Christian Jones wrote:> Hello! > I?m fitting a model with glm(family binomial). The best model counts 9 Variables and includes an interaction term that was generated by the product of to continuous variables (a*b). All variables are correlated under a value of 0.7 (Spearman rank order) While the estimates of both main effects are negativ, the resulting interaction term is positiv. This change of sign makes it difficult to interpret the model and above all, is this perhaps due to a bad variable choice ? > Thanks a lot for helpingRather than trying to interpret the model coefficients directly, you might visualize the a*b interaction effect. The effects package by John Fox is very useful for this: http://cran.r-project.org/doc/packages/effects.pdf> Christian > > _______________________________________________________________________ > Viren-Scan f?r Ihren PC! Jetzt f?r jeden. Sofort, online und kostenlos. > Gleich testen! http://www.pc-sicherheit.web.de/freescan/?mc=022222 > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code.-- Chuck Cleland, Ph.D. NDRI, Inc. 71 West 23rd Street, 8th floor New York, NY 10010 tel: (212) 845-4495 (Tu, Th) tel: (732) 512-0171 (M, W, F) fax: (917) 438-0894
Hello Chuck, many thanks for your advice. I?ll go through the package and see if I?ll manage to visualize the interaction effect this time as I already tried it with a special tool for displaying models with interaction terms in 3D. http://www.uni-oldenburg.de/landeco/Download/Software/LR_Mesh/LR_Mesh.htm but could not yield any plausible results. Hopefully this time... best regards Christian> -----Urspr?ngliche Nachricht----- > Von: Chuck Cleland <ccleland at optonline.net> > Gesendet: 31.08.06 14:16:12 > An: Christian Jones <ccatj at web.de> > CC: r-help at stat.math.ethz.ch > Betreff: Re: [R] need help with an interaction term> Christian Jones wrote: > > Hello! > > I???m fitting a model with glm(family binomial). The best model counts 9 Variables and includes an interaction term that was generated by the product of to continuous variables (a*b). All variables are correlated under a value of 0.7 (Spearman rank order) While the estimates of both main effects are negativ, the resulting interaction term is positiv. This change of sign makes it difficult to interpret the model and above all, is this perhaps due to a bad variable choice ? > > Thanks a lot for helping > > Rather than trying to interpret the model coefficients directly, you > might visualize the a*b interaction effect. The effects package by John > Fox is very useful for this: > > http://cran.r-project.org/doc/packages/effects.pdf > > > Christian > > > > _______________________________________________________________________ > > Viren-Scan f?r Ihren PC! Jetzt f?r jeden. Sofort, online und kostenlos. > > Gleich testen! http://www.pc-sicherheit.web.de/freescan/?mc=022222 > > > > ______________________________________________ > > R-help at stat.math.ethz.ch mailing list > > https://stat.ethz.ch/mailman/listinfo/r-help > > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > > and provide commented, minimal, self-contained, reproducible code. > > -- > Chuck Cleland, Ph.D. > NDRI, Inc. > 71 West 23rd Street, 8th floor > New York, NY 10010 > tel: (212) 845-4495 (Tu, Th) > tel: (732) 512-0171 (M, W, F) > fax: (917) 438-0894 >