Skip to content

extract the p value of F statistics from the lm class

5 messages · tedzzx, Thomas Petzoldt, Duncan Murdoch +1 more

tedzzx
# 
Dear R users

I have run an regression and want to extract the p value of the F
statistics, but I can find a way to do that.

x<-summary(lm(log(RV2)~log(IV.m),data=b))

Call:
lm(formula = log(RV2) ~ log(IV.m), data = b[[11]])

Residuals:
     Min       1Q   Median       3Q      Max 
-0.26511 -0.09718 -0.01326  0.11095  0.29777 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  -0.3059     0.1917  -1.595    0.121    
log(IV.m)     0.9038     0.1065   8.488 1.38e-09 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 

Residual standard error: 0.1435 on 31 degrees of freedom
Multiple R-squared: 0.6991,	Adjusted R-squared: 0.6894 
F-statistic: 72.04 on 1 and 31 DF,  p-value: 1.379e-09 

names(x)
 [1] "call"          "terms"         "residuals"    
 [4] "coefficients"  "aliased"       "sigma"        
 [7] "df"            "r.squared"     "adj.r.squared"
[10] "fstatistic"    "cov.unscaled"

x$fstatistic
   value    numdf    dendf 
72.04064  1.00000 31.00000 

But can not find the p value of F statistics. 

Thanks

Ted
Duncan Murdoch
#
On 05/04/2009 4:18 AM, tedzzx wrote:
If you're looking for something like that, the two places to look are:
  - the man page ?summary.lm (which gives the answer)
  - unclass(x) will display the object without the fancy printing, so 
you can see that the man page is accurate.  (Sometimes man pages are 
incomplete, and this way is needed, but not in this case.)

Duncan Murdoch
Thomas Petzoldt
# 
Hi,

what about the following:

## some test data
x <- 1:10
y <- x + rnorm(x)

## model and summary
m <- lm(y~x)
sm <- summary(m)
sm

# str(sm)
# sm$fstatistic

## and now: the manual case
1 - pf(sm$fstatistic[1], sm$fstatistic[2], sm$fstatistic[3])


Hope it helps, ThPe


tedzzx schrieb:
Duncan Murdoch
#
On 05/04/2009 4:18 AM, tedzzx wrote:
Sorry, I misread your question:  the p-value of that statistic isn't 
returned.  You just need to calculate it yourself, as

  f <- x$fstatistic
  pf(f[1], f[2], f[3], lower=FALSE)

Duncan Murdoch
(Ted Harding)
#
On 05-Apr-09 08:18:27, tedzzx wrote:
Maybe you were looking in the wrong place. A few lines above the
output from x$fstatistic

  x$fstatistic
     value    numdf    dendf 
  72.04064  1.00000 31.00000

you will find

  F-statistic: 72.04 on 1 and 31 DF,  p-value: 1.379e-09

and therefore will find the P-value. However, maybe that is not
the question you really wanted to ask. If that is what I think it
may be, you could

1:  Observe that x$fstatistic is a vector with 3 values which
    are: value of F; numerator df; demoninator df
2:  Note (from ?pf)
    pf(q, df1, df2, ncp, lower.tail = TRUE, log.p = FALSE)
3:  Therefore do
    pf(x$fstatistic[1],x$fstatistic[2],x$fstatistic[3],lower.tail=FALSE)
    # [1] 1.378626e-09

Note that the P-value is not in the list of values returned by lm()
although $fstatistic is one of the values. The computation of the
P-value in the displayed output from summary.lm() is done by the
'print' method for summary.lm() (just as in [3] above).

Hoping this helps,
Ted.

--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk>
Fax-to-email: +44 (0)870 094 0861
Date: 05-Apr-09                                       Time: 13:12:45
------------------------------ XFMail ------------------------------