# [R] mblm, confint problem

Kay Cichini Kay.Cichini at uibk.ac.at
Mon May 3 16:03:15 CEST 2010

```hello,

i did median based linear regression and computed conf.intervals for my
coefficients.
but something must have went wrong as the estimates are out of estimates
confidence bounds.
does someone know what is the problem here.

i get warning messages about wilcox.test is not able to do calculations of
exact p-values with ties and zeroes, but i do not expect the estimates and
theire ci's to be effected by this...

greetings,
kay

library(mblm)

y<-c(1.41421356237310, 2, 2.44948974278318, 0, 1.41421356237310,
2, 0, 1.41421356237310, 1.41421356237310, 1.73205080756888, 2,
1.41421356237310, 2.44948974278318, 1, 1.41421356237310, 3.16227766016838,
0, 1.41421356237310, 1.41421356237310, 1.41421356237310, 1, 1,
3, 1, 0)

x<-c(1.73205080756888, 2.23606797749979, 2.82842712474619,
1.41421356237310, 2.23606797749979, 2.44948974278318, 1, 1,
2.64575131106459,
2, 3.16227766016838, 2.23606797749979, 3.16227766016838, 2.64575131106459,
2.23606797749979, 3.16227766016838, 1, 1.73205080756888, 0,
1.73205080756888,
0, 1.41421356237310, 3.16227766016838, 0, 1)

mod<-mblm(y~x,repeated=F)
summary(mod)

mod\$coeff
confint(mod)

# gives:

> mod\$coeff
(Intercept)           x
0.3187684   0.6324555
>
> confint(mod)
0.025        0.975
(Intercept) -0.02332502 0.5550533
x                0.77481683 1.0266857          (?????)

-----
------------------------
Kay Cichini