i am try to use linear mixed effect model to explore the drivers (explanatory variables) of soil organic carbon (response variable), i was wondering if i can use cross validation to assess my model? how to plot the "observations "vs "predict value" ? is there some software can do that ? for example R, or using SAS.
P.S. I read a paper use SAS do that so, but i don't know how to calculate the estimate ?can anyone help me ?
the paper is in attachment. thank you .
Dear Daroui, I went through the paper only quickly. But I think that main idea is as follows - you make two groups of your data (say two thirds and one third). You build model from both. Then you compare the models (there are several ways of comparison - you can compare coefficients, you can compare model errors, etc., ideally the more comparisons, the better). Modelling can be done in any software such as SPSS or SAS.
hi jiri, thanks very much for your advice firstly. in my opinion he fit a model from 2 thirds of the data and use the left one third conducting a validation. I am now reading some books trying to fix the issue. best wishes.^^
There are two types of commonly used cross validation methods that you can use. You can refer to the textbook called elements of statistical learning by Hastie and Tibshirani. These are k fold cross validation and leave one out cross validation. Both have there advantages and disadvantages. There is freely available book called Introduction to Statistical learning which may have some examples on these based on R. I think you can also use the cvtools R package.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R