[STATSGRASS] sites import problem

[Roger Bivand]

On Mon, 17 Mar 2003, javier garcia - CEBAS wrote:

> Hi;
> I've imported a sites file into R, but get the following warning messages:
> ------------------------------------------------------------------------------------------------------
> > Cotas <- sites.get(G, slist="cotas_5000", debug=FALSE)
> > ls()
> [1] "Cotas"  "G"      "system"
> > summary(Cotas)
>       east            north              var1
>  Min.   :632206   Min.   :4163710   Min.   :   0.00
>  1st Qu.:657634   1st Qu.:4174570   1st Qu.:  65.13
>  Median :666414   Median :4179580   Median : 140.00
>  Mean   :    NA   Mean   :     NA   Mean   : 161.71
>  3rd Qu.:676889   3rd Qu.:4184840   3rd Qu.: 240.00
>  Max.   :688899   Max.   :4198600   Max.   :1116.18
> Warning messages:
> 1: Integer overflow in sum(.); use sum(as.numeric(.))
> 2: Integer overflow in sum(.); use sum(as.numeric(.))
> ---------------------------------------------------------------------------------------------
> I can't understand the consecuences of this warning nor how to avoid it
> Any help?

Which version of R, and the interface package are you using - it always 
helps to know?

If there are many points, then the sums of the coordinates may exceed 
INT_MAX=2147483647 in calculating the means. The warning means that if you 
need the mean of for example Cotas$east, which is represented as an 
integer, you could do summary(as.double(Cotas$east)), which will run.

For integer northings you run out of integers (at 4000km N of the equator)  
after 540 points, for your eastings you run out at about 3600 points.  
Another way to do this is to divide the coordinates by 1000 to represent
values as km.

I'll look at why sites.get() is returning integers, when it should 
probably be returning doubles.

Roger 

> 
> Javier
> 

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Breiviksveien 40, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 93 93
e-mail: Roger.Bivand at nhh.no