How to suppress correlation table in LME?

Question

In the standard example of the lme() function in the nlme package of R:

fm2 <- lme(distance ~ age + Sex, data = Orthodont, random = ~ 1)
summary(fm2)

there appears a correlation table:

Correlation: 
          (Intr) age   
age       -0.813       
SexFemale -0.372  0.000

which can be huge if there are many factor combinations involved.

Is there any way to suppress the output in the summary command? I know that I can use

   print(fm2, cor=F)

but this does not show me the rest of the usual output for example no p-value calculation.

score 4 · Accepted Answer · answered May 13 '13 at 12:13

4

Looking at nlme:::print.summary.lme I don't see a way to suppress the correlation matrix printing (although you could create a hacked version of that function removing the if clause beginning if (nrow(x$tTable)>1) ...)

Perhaps it would be useful to you to be able to print just the summary of the fixed-effect parameters ... ?

 printCoefmat(summary(fm2)$tTable)

answered May 13 '13 at 12:13

Ben Bolker

211,554
25
370
453

Thank you, that helps already (a bit). Sorry for the late upvote, I got distracted. – Jens Jun 23 '13 at 17:52

score 3 · Answer 2 · answered Jun 21 '13 at 03:51

3

Or, more concisely, summary(fm2)$tTable

answered Jun 21 '13 at 03:51

bokov

3,444
2
31
49

this works, but `printCoefmat()` makes the output prettier/more like the summary output ... – Ben Bolker Nov 29 '18 at 19:32

score 3 · Answer 3 · answered Aug 07 '15 at 13:33

I just recently ran into the same issue when fitting models with lots of fixed effects and the correlation table was huge and really cluttered up the output. Looking at print.summary.lme() (which isn't exported, so you have to use nlme:::print.summary.lme) shows that the part comes from these lines:

if (nrow(x$tTable) > 1) {
    corr <- x$corFixed
    class(corr) <- "correlation"
    print(corr, title = " Correlation:", ...)
}

as already pointed out by Ben. Instead of rewriting/replacing the entire function, we can also use a simple trick, replacing nlme:::print.correlation (which is what is actually doing the printing of the correlation matrix) with our own print method for objects of class correlation. This can be done with:

assignInNamespace("print.correlation", function(x, title) return(), ns="nlme")

Now the correlation matrix will be omitted, but you get the remaining output.

How to suppress correlation table in LME?

3 Answers3