0

My question is highly related to this one: R update() interaction term not dropped

However, I don't have multiple categories in my predictor variables, so I don't understand how my issue relates to the answer. Maybe I'm just not understanding it...

I'd like to remove the insignificant 3-way interaction terms in a model reduction process one at a time.

However, the following happens:

model1 <- lme(sum.leafmass ~ stand.td.Sept.2017*stand.wtd.Sept.2017*I((stand.td.Sept.2017)^2)*I((stand.wtd.Sept.2017)^2), random = ~1|block/fence, method="ML", data=subset(Total.CiPEHR, species=="EV"), na.action=na.omit)

model2 <- update(model1,.~.-stand.td.Sept.2017:stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2))
summary(model2)   ##works correctly to eliminate insignificant 4-way interactions

summary(model2)
                                                                         DF   t-value p-value
(Intercept)                                                               4  3.849259  0.0183
stand.td.Sept.2017                                                        4 -1.436666  0.2242
stand.wtd.Sept.2017                                                       4 -2.921806  0.0432
I((stand.td.Sept.2017)^2)                                                 4  4.594303  0.0101
I((stand.wtd.Sept.2017)^2)                                                4 -0.313197  0.7698
stand.td.Sept.2017:stand.wtd.Sept.2017                                    4 -1.301935  0.2629
stand.td.Sept.2017:I((stand.td.Sept.2017)^2)                              4  1.853451  0.1374
stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2)                             4  4.354757  0.0121
stand.td.Sept.2017:I((stand.wtd.Sept.2017)^2)                             4 -0.028199  0.9789
stand.wtd.Sept.2017:I((stand.wtd.Sept.2017)^2)                            4  1.598564  0.1852
I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2)                      4 -1.683214  0.1676
stand.td.Sept.2017:stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2)          4  1.972616  0.1198
stand.td.Sept.2017:stand.wtd.Sept.2017:I((stand.wtd.Sept.2017)^2)         4 -1.635314  0.1773
stand.td.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2)   4  2.190518  0.0936
stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2)  4 -0.968249  0.3877

##attempt to remove insignificant 3-way interaction
model3 <- update(model2,.~.,-stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2))
summary(model3)
                                                                         DF   t-value p-value
(Intercept)                                                               4  3.849259  0.0183
stand.td.Sept.2017                                                        4 -1.436666  0.2242
stand.wtd.Sept.2017                                                       4 -2.921806  0.0432
I((stand.td.Sept.2017)^2)                                                 4  4.594303  0.0101
I((stand.wtd.Sept.2017)^2)                                                4 -0.313197  0.7698
stand.td.Sept.2017:stand.wtd.Sept.2017                                    4 -1.301935  0.2629
stand.td.Sept.2017:I((stand.td.Sept.2017)^2)                              4  1.853451  0.1374
stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2)                             4  4.354757  0.0121
stand.td.Sept.2017:I((stand.wtd.Sept.2017)^2)                             4 -0.028199  0.9789
stand.wtd.Sept.2017:I((stand.wtd.Sept.2017)^2)                            4  1.598564  0.1852
I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2)                      4 -1.683214  0.1676
stand.td.Sept.2017:stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2)          4  1.972616  0.1198
stand.td.Sept.2017:stand.wtd.Sept.2017:I((stand.wtd.Sept.2017)^2)         4 -1.635314  0.1773
stand.td.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2)   4  2.190518  0.0936
stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2)  4 -0.968249  0.3877

##3-way interaction term still there.

Why won't the interaction term drop? The predictor variables are continuous and so should be independent from each other, right..?

Someone please explain if I'm not understanding something basic here...

Briana
  • 11
  • 1

1 Answers1

0

Solved my own question.

Dummy syntax error. (had an incorrect comma in the .~. portion)

###Incorrect syntax.
model3 <- update(model2,.~.,-stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2))

###Correct syntax.
model3 <- update(model2,.~.-stand.wtd.Sept.2017:I((stand.td.Sept.2017)^2):I((stand.wtd.Sept.2017)^2))
Briana
  • 11
  • 1