Gam with mrf smoother (mgcv)

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Gam with mrf smoother (mgcv)

gilla
Hallo,

I want to use gam from the mgcv package with a mrf smoother.

This is my data set (`x`)

        y          id
     1  0.6684496  1
     2  0.6684496  2
     3  0.6684496  3
     4  0.6684496  4
     5  0.6684496  5
     6  0.6684496  6
     7  0.6684496  7
     8  0.5879492  8
     9  0.5879492  9
     10 0.5879492 10
     11 0.5879492 11
     12 0.5879492 12
     13 0.5879492 13
     14 0.5879492 14
     15 0.6445448 15
     16 0.6445448 16
     17 0.6445448 17
     18 0.6445448 18
     19 0.6445448 19
     20 0.6445448 20
     21 0.6445448 21
     22 0.5815767 22
     23 0.5815767 23
     24 0.5815767 24
     25 0.5815767 25
     26 0.5815767 26
     27 0.5815767 27
     28 0.5815767 28
     29 0.6636673 29
     30 0.6636673 30
     31 0.6636673 31
     32 0.6636673 32
     33 0.6636673 33
     34 0.6636673 34
     35 0.6636673 35
     36 0.4249320 36
     37 0.4249320 37
     38 0.4249320 38
     39 0.4249320 39
     40 0.4249320 40
     41 0.4249320 41
     42 0.4249320 42

and this is my neighbouring structure (nb)

     $`1`
     [1] 1 2 3 4 5 6 7

     $`2`
     [1] 1 2 3 4 5 6 7

     $`3`
     [1] 1 2 3 4 5 6 7

     $`4`
     [1] 1 2 3 4 5 6 7

     $`5`
     [1] 1 2 3 4 5 6 7

     $`6`
     [1] 1 2 3 4 5 6 7

     $`7`
     [1] 1 2 3 4 5 6 7

     $`8`
     [1]  8  9 10 11 12 13 14

     $`9`
     [1]  8  9 10 11 12 13 14

     $`10`
     [1]  8  9 10 11 12 13 14

     $`11`
     [1]  8  9 10 11 12 13 14

     $`12`
     [1]  8  9 10 11 12 13 14

     $`13`
     [1]  8  9 10 11 12 13 14

     $`14`
     [1]  8  9 10 11 12 13 14

     $`15`
     [1] 15 16 17 18 19 20 21

     $`16`
     [1] 15 16 17 18 19 20 21

     $`17`
     [1] 15 16 17 18 19 20 21

     $`18`
     [1] 15 16 17 18 19 20 21

     $`19`
     [1] 15 16 17 18 19 20 21

     $`20`
     [1] 15 16 17 18 19 20 21

     $`21`
     [1] 15 16 17 18 19 20 21

     $`22`
     [1] 22 23 24 25 26 27 28

     $`23`
     [1] 22 23 24 25 26 27 28

     $`24`
     [1] 22 23 24 25 26 27 28

     $`25`
     [1] 22 23 24 25 26 27 28

     $`26`
     [1] 22 23 24 25 26 27 28

     $`27`
     [1] 22 23 24 25 26 27 28

     $`28`
     [1] 22 23 24 25 26 27 28

     $`29`
     [1] 29 30 31 32 33 34 35

     $`30`
     [1] 29 30 31 32 33 34 35

     $`31`
     [1] 29 30 31 32 33 34 35

     $`32`
     [1] 29 30 31 32 33 34 35

     $`33`
     [1] 29 30 31 32 33 34 35

     $`34`
     [1] 29 30 31 32 33 34 35

     $`35`
     [1] 29 30 31 32 33 34 35

     $`36`
     [1] 36 37 38 39 40 41 42

     $`37`
     [1] 36 37 38 39 40 41 42

     $`38`
     [1] 36 37 38 39 40 41 42

     $`39`
     [1] 36 37 38 39 40 41 42

     $`40`
     [1] 36 37 38 39 40 41 42

     $`41`
     [1] 36 37 38 39 40 41 42

     $`42`
     [1] 36 37 38 39 40 41 42

However, when I run

     b <- gam(y ~ s(id,bs="mrf",xt=list(nb=nb)),data=x,method="REML")

I get the following error:

Error in initial.sp(w * x, S, off) : S[[1]] matrix is not +ve definite.

Why is the penalty matrix non semi-positive definite? Am I defining
something wrong?

Many thanks in advance

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.