[Gstat-info] cross validation report

[Edzer pebesma]

no, it's the linear correlation coefficient.

pred-obs is prediction minus observed, pred.std. is prediction standard 
error, zscore is residual divided by pred.std.

hth,
--
Edzer

andy schrieb:
> Dear all,
> it's my first post in this list, and I want to say thank you to all 
> gstat people. Really a good application.
>
> I'm using Gstat through Idrisi Andes. I'm working with kriging and I 
> produce a cross validation report like that one I'm attaching.
>
> I think it's a standard gstat cross validation report. If I'm right, I 
> need to know the meaning of these parameters:
> - corr(Obs, Pred)
> - pred.-obs.
> - pred.std.
> - zscore
>
> I think "corr(Obs, Pred)" is correspondence. How is it computed?
>
> Thank you,
>
> Andrea
>
> _________________________________________________________________________________________ 
>
>
> corr(Obs, Pred): 0.2586  [using ordinary kriging]
>
>              observed   predicted   pred.-obs.   pred.std.     zscore
> ======================================================================
> minimum          39.53       429.6       -9783        2516      -3.641
> 1st q.           529.4        1183       -1419        2545     -0.5321
> median            1578        2085       552.8        2593      0.2069
> 3rd q.            3914        3307        1701        2654      0.6537
> maximum      1.206e+04        5656        3491        2856       1.321
>
> n                   28          37          28          37          28
> mean              2975        2280      -529.1        2612     -0.2077
> std.dev.          3461        1313        3383       83.13       1.286
> _______________________________________________
> Gstat-info mailing list
> Gstat-info at geo.uu.nl
> http://mailman.geo.uu.nl/mailman/listinfo/gstat-info