[R] t-values for two-way interactions

R. Baker rb373 at cam.ac.uk
Thu Jul 5 11:58:50 CEST 2007

I have a model with 3 fixed factors (type, stress, MorD) and two 
significant two-way interactions (type*stress, stress*MorD).


#                  Estimate Std.Error  DF t.value pvals ci950 ci990 ci999

#(Intercept)        241.738     8.757 994  27.606 0e+00  TRUE  TRUE  TRUE

#typePsPr           -26.516     5.905 994  -4.490 1e-05  TRUE  TRUE  TRUE

#stressPN           -21.820     1.776 994 -12.284 0e+00  TRUE  TRUE  TRUE

#MorDMis            -26.588     5.905 994  -4.503 1e-05  TRUE  TRUE  TRUE

#typePsPr:stressPN   13.904     2.051 994   6.779 0e+00  TRUE  TRUE  TRUE

#stressPN:MorDMis     9.168     2.051 994   4.470 1e-05  TRUE  TRUE  TRUE

I would like to find out more about the type*stress interaction, e.g. how 
the levels of type differ between stress. More specifically I want to find 
a t-value for the pairwise comparisons:

typePr:stressN vs. typePsPr:stressN
typePr:stressPN vs. typePsPr:stressPN

I can't do this with the summary table (above) as it always includes the 
factor MorD (mis/dis) in each pairwise comparison, i.e. it provides 
t-values for

typePr:stressN:MorDmis and typePsPr:stressN:MorDmis 
typePr:stressN:MorDdis and typePsPr:stressN:MorDdis 

Is it possible to look at interaction of type*stress over MorD and NOT 
within MorDmis or MorDdis?

Any advice would be greatly appreciated!

Rachel Baker

PhD student                
Dept of Linguistics        
Sidgwick Avenue
University of Cambridge              

More information about the R-help mailing list