R help - Sep 2005 - I have a problem with npmc

Hi to all:

The reason for my mail is this:

I am tryng to make an analysis with non-parametric multiple comparisons with a
data set from Pinciples and Procedures of Satistics (Steel and Torrie,1980). The
data are of a one-way layouth:

 > DOKNOPAR
   class  var
1   DOK1 19.4
2   DOK1 32.6
3   DOK1 27.0
4   DOK1 32.1
5   DOK1 33.0
6   DOK5 17.7
7   DOK5 24.8
8   DOK5 27.9
9   DOK5 25.2
10  DOK5 24.3
11  DOK4 17.0
12  DOK4 19.4
13  DOK4  9.1
14  DOK4 11.9
15  DOK4 15.8
16  DOK7 20.7
17  DOK7 21.0
18  DOK7 20.5
19  DOK7 18.8
20  DOK7 18.6
21 DOK13 14.3
22 DOK13 14.4
23 DOK13 11.8
24 DOK13 11.6
25 DOK13 14.2
26   COM 17.3
27   COM 19.4
28   COM 19.1
29   COM 16.9
30   COM 20.8

But when i try making de npmc, 
> npmc(DOKNOPAR)
R displays this message:



Error in uniroot(f = function(arg) p - z.dist(arg, corr = corr, df = df,  : 
        f() values at end points not of opposite sign

What is the reason for this? How can i make this npmc analysis correctly?

Daniel Rodriguez

Universidad Nacional de Colombia

Daniel Rodriguez Caicedo <drodriguezca <at> unal.edu.co> writes:
> 
> 
> I am tryng to make an analysis with non-parametric multiple comparisons
with
a data set from Pinciples and
> Procedures of Satistics (Steel and Torrie,1980). The data are of a one-way 
layouth:
> 
>  > DOKNOPAR
>    class  var
> 1   DOK1 19.4
> 2   DOK1 32.6
> 3   DOK1 27.0
...
> 
> But when i try making de npmc, 
> 
> > npmc(DOKNOPAR)
> 
> R displays this message:
> 
> Error in uniroot(f = function(arg) p - z.dist(arg, corr = corr, df = df,  :
>         f() values at end points not of opposite sign
> 
....

Works for me under Windows, current R and packages, but needs some time. Try a 
str(DOKNOPAR), maybe your data are not exactly what you expecte

   cmp gn effect      lower.cl     upper.cl ...
1  1-2 10   0.94  0.5811513117 1.2988486883 ...
2  1-3 10   0.00 -0.0001573656 0.0001573656 
3  1-4 10   0.18 -0.5714102497 0.9314102497 
4  1-5 10   0.88  0.2349931421 1.5250068579 
....

Dieter