1

My apologies for not including enough information to properly access the issue. Anyway,I want to compare different pairs of linear models and record their p-values into a table. Where my troubles lie is that I cannot extract only the p-value.

TC <- matrix(data=NA, nrow=1, ncol=6)
    ML5 <- 1:5
for(B in ML5) {
    Coop <- anova(M6,Models5[[B]])$"Pr(>F)"
    TC[1,B] <- Coop
}

 >Coop
   NA 2.041767e-05

So how do I only put the number into the table and not the NA into the table?

Thanks a million!

Lcat91
  • 804
  • 3
  • 10
  • 14

3 Answers3

7

You just need quotation marks:

pvalue <- anova(m2,m1)$"Pr(>F)"

You can access the second element of pvalue using normal bracket subsetting:

pvalue[2]

So in your example I believe you'll use 

Coop <- anova(M6,Models5[[B]])$"Pr(>F)"[2]

(Although without access to M6, I can't be totally sure).

The str command is very useful in this sort of situation, to figure out what kind of object you're dealing with: 

str(myanova$"Pr(>F)")
Drew Steen
  • 16,045
  • 12
  • 62
  • 90
  • OK now I get the error, "Error in T21Comp[A, B] <- Coop : number of items to replace is not a multiple of replacement length" – Lcat91 Aug 07 '13 at 19:55
  • Hard to interpret that without a reproducible example. – Drew Steen Aug 07 '13 at 20:27
  • `myanova$"Pr(>F)"` is a vector, not a single number. People get caught by surprise because the `NA` here are invisible in the printout of an intact `anova` object (which is really just a `data.frame`), but still exist and become visible when subset. The error probably means you are assigning it to something that has fewer spaces than `Coop` takes up. Print out `T21Comp[A,B]` and `Coop` (by typing them at the command line) and see how big each one is. Subset the larger one accordingly (or change the length of the smaller one). – bokov Aug 09 '13 at 23:28
0

Using mtcars data from R (I would prefer the solution suggested by @Drew Steen)

myanova<-anova(lm(mpg~cyl,mtcars),lm(mpg~cyl+disp,mtcars))
pval<-summary(myanova)[1,6]

Updated as per question: 

models1<-lm(mpg~cyl,mtcars)
models2<-list(lm(mpg~cyl+disp,mtcars),lm(mpg~cyl+disp+hp,mtcars),lm(mpg~cyl+disp+hp+drat,mtcars))
myp1<-lapply(models2,function(x) anova(models1,x))
myp2<-lapply(myp1,function(x)summary(x)[1,6])
> myp3
[1] "   0.05419  " "   0.09891  " "   0.09661  "

Add:If you are using "$"Pr(>F)"

  myp2<-lapply(myp1,function(x)x$"Pr(>F)")
  myp3<-sapply(myp2,function(x) as.data.frame(x)[2,1])
    > myp3
    [1] 0.05418572 0.09891469 0.09661222
Metrics
  • 15,172
  • 7
  • 54
  • 83
0

Using aov():

anova_a <- aov(response ~ time + Error(subject), data = all_data)

f_value <- unlist(summary(anova_a)[[2]])[9]
Ian
  • 1