Design Matrix -- Scaling Covariates

Warning. You should scale the values of covariates used in the Design Matrix to have a mean in the interval [0-1] to ensure that the numerical optimization algorithm finds the correct parameter estimates. For example, suppose the variable Harvest, measured in thousands of animals with a range from 200 to 9000, is used as a covariate in the Design Matrix. By entering the values of Harvest as 0.02 to 0.9 instead of 200 to 9000, you will find that the numerical optimization procedure performs better. Note that in this example, each value of Harvest was multiplied by 0.0001 to obtain the value entered in the Design Matrix. To correct the beta estimate of harvest from this model to use with the original harvest values, you would multiply the value by 0.0001. Likewise, the standard error of this beta estimate would be multiplied by 0.0001 to obtain the standard error for the beta value to use with the uncorrected harvest estimates.

Individual covariates should also be standardized, but MARK can do this automatically.