Dear John Fox and everyone,
I have been using the effects library with glms and have found it very
useful.
Now I'm trying it with lms and I'm not sure if the results of the
allEffects() are as expected.
I've got a model that looks like this:
mymodel = lm(formula = A ~ B + C + D + B:D + C:D)
Residuals:
Min 1Q Median 3Q Max
-3.80156 -0.73486 -0.09792 0.63602 4.77747
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.551727 0.014788 37.310 < 2e-16 ***
B 0.112033 0.014067 7.964 1.82e-15 ***
C 0.150992 0.010281 14.686 < 2e-16 ***
D 0.319938 0.018451 17.340 < 2e-16 ***
B:D 0.042949 0.008208 5.233 1.70e-07 ***
C:D 0.077968 0.010054 7.755 9.58e-15 ***
---
Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1
Residual standard error: 1.083 on 11555 degrees of freedom
Multiple R-squared: 0.2412, Adjusted R-squared: 0.2409
F-statistic: 734.6 on 5 and 11555 DF, p-value: < 2.2e-16
When I ran plot(allEffects(mymodel)), I was expecting that the effect
diagrams for the interaction terms will include the effects of the
individual terms as well. However, this was not the case, since despite the
strong individual terms, with one of the terms set to zero the regression
line for the second one was invariably A=0.
I'm wondering if this is a bug or a "feature" and in the latter
case, what
can be done to display effects of the terms in a "combined" way, such
that
the contribution of both them on their own and as part of an interaction
term is taken into account.
Many thanks
Mikhail
--
Mikhail Spivakov, PhD
European Bioinformatics Institute
Hinxton
Cambridgeshire CB10 1SD
UK
spivakov at ebi.ac.uk
--
View this message in context:
http://www.nabble.com/allEffects%28%29-with-lm-tp23837461p23837461.html
Sent from the R help mailing list archive at Nabble.com.
