SAS code covariance structure formula for spherical, gaussian and power in proc mixed

[TA2020]

Hi, 

I need to compute the covariance, semivariogram. I used spatial spherical, spatial gaussian and spatial power structure. I found the formula but having a problem to write the code based on these formula correctly. Please anyone suggest me some documents where I can see how to apply those equations.

The formula are in the following link:

 https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.3/statug/statug_variogram_details02.htm

TIA

[sbxkoenk]

Hello,

Are you asking on how to work with non-supported covariance structures in PROC MIXED??
That's not possible as far as I know!

See here :

Usage Note 37107: Comparing covariance structures in PROC MIXED
https://support.sas.com/kb/37/107.html

Here's the list of supported R-side covariance structures :
https://go.documentation.sas.com/doc/en/statug/15.3/statug_mixed_syntax14.htm#statug.mixed.repeateds...

If you give a bit more info, we might be able to help you out ...

Cheers,

Koen

[SteveDenham]

I'll follow up on @sbxkoenk 's comment - what are you trying to do here? PROC MIXED will give you parameter estimates for spatial covariances based off the residuals, after fitting the fixed effects. PROC VARIOGRAM will give you parameter estimates for spatial covariance structures based on the observed values, with only the location as an effect.

In any case, the formulae for the various structures are all in the documentation.

SteveDenham

[TA2020]

Hi, Steven

Can we compute variogram with proc mixed code? I'm trying to try it with proc mixed 

Thanks

[TA2020]

Hi, sbxkoenk

I'm sorry for replying this late. I forgot my password, and kept busy looking for more documentation.

Thanks for your suggestion. I actually meant the covariance structure formula shown in table 18. for Spatial Covariance structure (your 2nd link) .  My data after log transformation fitted for mixed model. I have three years but small dataset (in total 119), year as fixed effect.  Now I am looking at the scaled residuals and comparing it between selected spatial covariance structure and independence model. Some of variables are correlated to each other, so I expected that independence model could not be best fit. I am using proc mixed for calculating correlogram, covariogram and semivariance using the formula given in table 18 (R-side covariance structure). I'm using SP Gaussian and SP Power spatial covariance structure.

 Do you know about any general formula for estimating semivariance from residuals, variences and covariance?

I'm actually new in Geostatistics analysis. I might get misconception about it. I have been working on SAS from 2022. I still need to explore more in SAS for spatial analysis. 

Thanks