Comparing fitted predmicror models

This vignette shows a compact workflow for comparing alternative predictive microbiology models with the predmicror_fit interface.

Fit candidate models

library(predmicror)

data(growthfull)

huang <- fit_growth(
  growthfull,
  model = "HuangFM",
  time = "Time",
  response = "lnN",
  start = list(Y0 = 0, Ymax = 22, MUmax = 1.7, lag = 5)
)

baranyi <- fit_growth(
  growthfull,
  model = "BaranyiFM",
  time = "Time",
  response = "lnN",
  start = list(Y0 = 0, Ymax = 22, MUmax = 1.7, lag = 5)
)

Extract fitted values and residuals

head(predmicror_augment(huang))

The returned data frame keeps the original columns and adds:

.fitted: predicted response on the fitted response scale;
.resid: observed minus fitted response, when the response column is available;
.model: fitted model name;
.type: model family.

Calculate diagnostics for one model

fit_metrics(huang)

fit_metrics() reports residual and information-criterion diagnostics on the response scale used for fitting.

Compare models

compare_models(
  huang = huang,
  baranyi = baranyi,
  sort_by = "AIC"
)

For models fitted to the same response variable and dataset, lower AIC, BIC, RMSE, and MAE values usually indicate a better fit. These criteria should be interpreted together with residual plots, biological plausibility, and parameter uncertainty.