The indescr option in WOMBAT is used to specify a residual covariance structure between all observations for different traits within an individual. This is particularly useful when you suspect that the residual errors for different traits are correlated.

Data File Format for indescr

While WOMBAT doesn't have a specific data file format for the indescr option, it's typically handled within the parameter file. The parameter file is where you specify the model, fixed and random effects, and other options, including the residual covariance structure.

Example Parameter File with indescr

Here's a simplified example of a parameter file that uses the indescr option:

**DATA** FILE = mydata.dat FORMAT = FREE MISSING = -99 **VARIABLE** ANIMAL = 1 TRAIT1 = 2 TRAIT2 = 3 **MODEL** MODEL = ANIMAL **RANDOM** ANIMAL = PED **COVARIANCE** INDESCR

Explanation:

Additional Considerations:

• Pedigree File: A pedigree file is required for the animal model, specifying the relationships between animals.
• Model Complexity: The indescr option can be combined with other covariance structures (e.g., genetic covariance, environmental covariance) to fit more complex models.
• Convergence: Fitting models with the indescr option can be computationally intensive, especially for large datasets. It's important to monitor convergence and consider alternative approaches if necessary.

Assistance

If you're struggling with the specific implementation of the indescr option or have questions about your data and model, consider the following:

By carefully specifying the indescr option in your parameter file and ensuring your data is formatted correctly, you can effectively model the residual covariance structure between traits in WOMBAT.

Dear Adnan,

Thank you for your guidance and assistance. I managed to sort out data file. The only remaining issue is that when I fit the model through parameter file: I have trait 1 with 8 observation for 4 months (meaning 8 observations every month), I also have trait 2 with 30 observation every month for 4month period. When I try to fit the covariance in correlation between traits 1 and traits 2, trait 1 with small number of observation tends to increase its raw variance...when I change it in the parameter file, the wombat tells me that the residual variance went drastically wrong and stops to compute the genetic and phenotypic correlation between the traits. Any advice on how to set the covariance matrix?

Thank you once again.