[STATSGRASS] gstat - trend over 3th polinominal

[Jarek Jasiewicz]

Edzer J. Pebesma napisał(a):
> Jarek, we don't know what your "newdata" object is. It is a name of an 
> argument, but passed as data; I would expect something like
>
> predict(g, newdata = meuse.grid)
yes, it was an erorr during retyping command of course newdata=meuse.grid
>
> I didn't know of that form in formulas, so thanks! Pretty much shorter 
> than writing the 4th order polynomial out.
I'm not shure if it really works (i.e it is mathematically proper) - I 
only point formula which works
>
> It might work "in principle" but you may expect increasing 
> inaccuracies with higher order trends due to the large difference 
> between say x^4 and x. Rescaling the coordinates prior to the analysis 
> (e.g. dividing by 1000, going from m to km) might help. poly() doesn't 
> help here, as it will rescale differently for meuse and meuse.grid, 
> afaik.
>
> I also wonder what, besides curiosity, the value of a 4-th or higher 
> order trend surface might be.
> -- 
> Edzer
>
> Jarek Jasiewicz wrote:
>> Hi
>> I found that result of formula:
>>
>> gs=gstat(id="x", formula=zinc~(x+y)^2, data=meuse)
>> pgs=predict(gs, newdata, meuse.grid)
>>
>> is equvalent to:
>>
>>
>> gs=gstat(id="x", formula=zinc~x+y, data=meuse, degree=2)
>> pgs=predict(gs, newdata, meuse.grid)
>>
>>
>> so I tried:
>> gs=gstat(id="x", formula=zinc~(x+y)^4, data=meuse)
>> pgs=predict(gs, newdata, meuse.grid)
>>
>> and it worked!
>>
>> so it is possible to calculate trend over 3th polynominal (now its 
>> not a metter for what :))) in that way?
>>
>> regards
>> Jarek
>>
>>
>>
>> _______________________________________________
>> statsgrass mailing list
>> statsgrass at grass.itc.it
>> http://grass.itc.it/mailman/listinfo/statsgrass