Output files #7
Description
I am refactoring how PMcore writes output files to make it more modular.
In this process, I am suggesting we change some of the output files.
Current files
settings.json
Contains the settings, parameter boundaries, etc. for the run.
cycles.csv
This file contains the cycle number, convergence status, status for the algorithm, the objective function value (-2 * LL), the number of support points, and the values of gamma/lambda for each error model (i.e. output equation).
covs.csv
The (interpolated) covariate value at each observation time.
obs.csv
For each subject and output equation, the observed concentration.
op.csv
Like op.csv, but contains the predictions for each observation.
pred.csv
Like op.csv, but does not contain the observations, and contains predictions made at each idelta.
theta.csv
Contains the support point parameter values and (population) probabilities.
posterior.csv
The long format of theta.csv, where the prob column is replaced with the posterior probabilities, and the new point column refers to the (zero-)index of the support point.
Suggested changes
obs.csv
Contains all observations AND the (interpolated) covariate values for each observation.
Consequence: Eliminates the need for covs.csv
pred.csv
Combines obs.csv, op.csv, and pred.csv. In other words, we add the observations to the current pred.csv. When predictions are made at idelta intervals, the observation is NA or missing.
We could also add the interpolated covariate values to this file. I can see this being useful for pharmacodynamic models.
Consequence: Eliminates the need for op.csv