In Simbiology, how to find features (covariates) which will increase the loglikelihood greater than the base model.
1 view (last 30 days)
In population pharmacokinetic analysis, suppose that I have finalised a base model (after exploring all the available model), and let us say keep the base model as non-linear mixed effect model with exponent error with FOCE method of loglikelihood estimation in a two compartmental PK model.
The base model was selected on the basis of highest loglikelihood value. Base model is defined as the model which contains only the PK parameters without the covariates and the PK parameters are defined according to standard two compartmental guidelines. I would like to increase the loglikelihood by means of adding the covariate into the context (that is the essence of population pharmacokinetics).
Is there any tool in Simbiology, which would help to find the features (covariates) which will increase the loglikelihood greater than the base model. I know that it can be done manually one by one. But is there a fast trick for getting a good initial guess.
On continuation of the same question, how much increase in loglikelihood would i consider inclusion of the covariate in the modelling. The consensus of the literature says that decrease by a value of 6 in Objective Function Value warrants inclusion of the covariate in the model. Objection Function Value is defined as - 2 times the loglikelihood ; and the interpretation of Objective Function Value will be different, a decrease of Objective Function Value is better