Quantcast

Appropriate variogram model

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
3 messages Options
Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Appropriate variogram model

R-sig-geo mailing list
Hi,

Can anyone suggest me the most appropriate variogram model to fit my data? 

I have tried three different models and results are too bad. Here is the comparison of different variogram models as attached.

Thanks,
Ankur

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo

Different models.pdf (38K) Download Attachment
Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Appropriate variogram model

Dr. Benedikt Gräler
Dear Ankur,

I see two issues from your plot:
i) there seems to be a seasonality in your data (maybe daily?) as the
sample variogram grows up to half the length of your temporal axis and
then drops again
ii) the starting values of your fitting routine are far off from what
can be deduced from the sample variogram. Hence, the numerical routines
have a hard time to identify the correct values. Additionally, the
parameters are of different orders of magnitude further complicating the
numerical fit. Manual re-scaling might help.

Try adding the parameter "scales=list(arrows=F)" for a visual assessment
of the empirical variogram; in your case

plot(var, wireframe=T, scales=list(arrows=F))

This reveals that the variogram indicates the same strength of
dependence for values 24 and 0 hours apart -> daily cycle? I'd suggest
to model the daily cycle first and then to model the spatio-temporal
variogram of the residuals.

HTH,

  Ben



On 08/05/2017 06:33, Ankur Sarker via R-sig-Geo wrote:

> Hi,
>
> Can anyone suggest me the most appropriate variogram model to fit my data?
>
> I have tried three different models and results are too bad. Here is the
> comparison of different variogram models as attached.
>
> Thanks,
> Ankur
>
>
> _______________________________________________
> R-sig-Geo mailing list
> [hidden email]
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>
--
Dr. Benedikt Gräler
52°North Initiative for Geospatial Open Source Software GmbH
Martin-Luther-King-Weg 24
48155 Muenster, Germany

E-Mail: [hidden email]
Fon: +49-(0)-251/396371-39
Fax: +49-(0)-251/396371-11

http://52north.org/
Twitter: @FiveTwoN

General Managers: Dr. Albert Remke, Dr. Andreas Wytzisk
Local Court Muenster HRB 10849

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo

b_graeler.vcf (435 bytes) Download Attachment
Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Appropriate variogram model

R-sig-geo mailing list
Dear Dr. Benedikt Gräler,

Thank you for your advice. Now, I have found some directs to move on. 

Actually, I do not have enough background to analyze by myself. Based on your suggestions, I have plotted the variogram. I think there are daily and weekly seasonality.

According to your suggestions, I am trying to model daily seasonality. It would be great if you can give me more advice.

With Best Regards,
Ankur


On Wednesday, May 17, 2017 6:10 AM, Dr. Benedikt Gräler <[hidden email]> wrote:


Dear Ankur,

I see two issues from your plot:
i) there seems to be a seasonality in your data (maybe daily?) as the
sample variogram grows up to half the length of your temporal axis and
then drops again
ii) the starting values of your fitting routine are far off from what
can be deduced from the sample variogram. Hence, the numerical routines
have a hard time to identify the correct values. Additionally, the
parameters are of different orders of magnitude further complicating the
numerical fit. Manual re-scaling might help.

Try adding the parameter "scales=list(arrows=F)" for a visual assessment
of the empirical variogram; in your case

plot(var, wireframe=T, scales=list(arrows=F))

This reveals that the variogram indicates the same strength of
dependence for values 24 and 0 hours apart -> daily cycle? I'd suggest
to model the daily cycle first and then to model the spatio-temporal
variogram of the residuals.

HTH,

  Ben



On 08/05/2017 06:33, Ankur Sarker via R-sig-Geo wrote:

> Hi,
>
> Can anyone suggest me the most appropriate variogram model to fit my data?
>
> I have tried three different models and results are too bad. Here is the
> comparison of different variogram models as attached.
>
> Thanks,
> Ankur
>
>
> _______________________________________________
> R-sig-Geo mailing list
> [hidden email]
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo

>

--
Dr. Benedikt Gräler
52°North Initiative for Geospatial Open Source Software GmbH
Martin-Luther-King-Weg 24
48155 Muenster, Germany

E-Mail: [hidden email]
Fon: +49-(0)-251/396371-39
Fax: +49-(0)-251/396371-11

http://52north.org/
Twitter: @FiveTwoN

General Managers: Dr. Albert Remke, Dr. Andreas Wytzisk
Local Court Muenster HRB 10849

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo



_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo

variogram.pdf (17K) Download Attachment
Loading...