Displaying 2 results from an estimated 2 matches for "lmall".
Did you mean:
small
2003 Jul 16
1
step.lm() fails to drop {many empty 2-way factor cells} (PR#3491)
...teractions have empty cells, e.g.
with(Dat, table(f1,f4)) #!
with(Dat, table(f1,f3))
## see the full design:
with(Dat, ftable(table(f1,f2,f3,f4))) ##--> many many 0
pairs(Dat, col = 1+as.integer(Dat$f1), pch = 1+as.integer(Dat$sex))
## Fit a "full model" with all 2-way interactions :
lmAll <- lm(formula = y ~ .^2, data = Dat)
summary(lmAll) # many insignificant ones {BTW: why is it bad to have "*" here?}
drop1(lmAll)
## clearly shows interaction terms that should be dropped (would decrease AIC)
## BUT:
sr <- step(lmAll)
## nothing eliminated ?step claims it would w...
2003 Jul 21
0
step.lm() fails to drop {many empty 2-way factor cells} (PR#3527)
...at, table(f1,f4)) #!
> with(Dat, table(f1,f3))
>
> ## see the full design:
> with(Dat, ftable(table(f1,f2,f3,f4))) ##--> many many 0
> pairs(Dat, col = 1+as.integer(Dat$f1), pch = 1+as.integer(Dat$sex))
>
> ## Fit a "full model" with all 2-way interactions :
> lmAll <- lm(formula = y ~ .^2, data = Dat)
> summary(lmAll) # many insignificant ones {BTW: why is it bad to have "*" here?}
>
> drop1(lmAll)
> ## clearly shows interaction terms that should be dropped (would decrease AIC)
>
> ## BUT:
> sr <- step(lmAll)
> ## not...