[R-sig-Geo] difference in coeeficient variance calculated from the ANOVA table and from the imat covariance matrix

Rachel Chocron rachel.saf at mail.huji.ac.il
Thu Nov 20 11:31:37 CET 2014


Just a small correction, I did *not *square root the std. error, I raised
it by power of 2.

2014-11-20 10:22 GMT+02:00 Rachel Chocron <rachel.saf at mail.huji.ac.il>:

> Hello!
> I used the function spdep.spautolm to calculate spatial regression:
>
> *text="arsin_species_num_proportion  ~
> I(POINT_X/100000)+I((POINT_X/100000)^2)+I(mean_400m/1000)+I((mean_400m/1000)^2)+I(summer_temperature)+I(sqr_temperature)+I(year_precipitation/1000)+I((year_precipitation/1000)^2)
> +I(RANGE_400m/1000)+I((RANGE_400m/1000)^2)+LandCoverNum_400m+sqr_LandCoverNum_400m"*
>
> *model400mSar <-spautolm(formula = as.formula(text), data = pointsDF,
> listw = weightsb,  family = "SAR", method = "eigen", verbose = TRUE,
> zero.policy = TRUE)*
> *Then I wanted to retrieve the coefficient covariance matrix.*
>
> So I did it using the imat object:
> *model400mSar$fit$imat*
>
> and I compared the variance of each coefficient according to the imat
> object,
> with the variance calculated from the ANOVA table extracted by:
>
> *summary(model400mSar)$Coef*
>
> and by square root the std. error of each coefficient
>
> and the results were different.
>
> I also tried the fdHess covarianve matrix but it also gave different
> results.
>
> *model400mSar$fdHess*
>
> My question is:
> How can I retrieve the covariance matrix of the coeeficients that is
> presented in the ANOVA table?
>
> Please see attached word documents with the fdHess matrix, the imat matrix
> and the ANOVA table.
>
>
> Many thanks,
> Rachel
>

	[[alternative HTML version deleted]]



More information about the R-sig-Geo mailing list