Models_boot: The list of the models refitted on bootstrapped data

In courtiol/SileR: Survival Analysis of Elephants Using an Extension of the William Siler's Framework

Description

This dataset contains the 40 bootstraps replicate for each of the 17 models. These models have been fitted on the complete dataset!

Format

A list with the 680 outputs of the function CI_all_models.

References

This study.

Examples

## The names of the models:
names(Models_boot)

## The AIC of model 1 replicate 1:
Models_boot[["mod1_1"]]$AIC

## The convergence status of model 1 replicate 1:
Models_boot[["mod1_1"]]$message

## The estimates from model 1 replicate 1:
data.frame(estimates = Models_boot[["mod1_1"]]$param_matrix[, "value"])

## Studying the effect of potential convergence issues:
table(unlist(lapply(Models_boot, function(mod) mod$convergence))) ## 1 = perfect convergence
models_with_pb <- which(unlist(lapply(Models_boot, function(mod) mod$convergence))!=1)
CI_with_pb <- table_model_averaging(models_list = Models,
                                    models_list_boot = Models_boot)
CI_without_pb <- table_model_averaging(models_list = Models,
 models_list_boot = Models_boot[!(names(Models_boot) %in% names(models_with_pb))])
max(abs(100*((CI_with_pb$SE - CI_without_pb$SE) /  CI_without_pb$SE)), na.rm = TRUE)
### Conclusion: only 3 fits out of 680 reached maximal time allowed before
### reaching full convergence. Those 3 model fits impact the SE of parameter
### estimates by less than 1.8% in the worst situation, which is negligible.
### We thus chose not to exclude them and base all CI computation on all 680
### fits.

courtiol/SileR documentation built on May 16, 2020, 8:10 p.m.