Skip to content

Instantly share code, notes, and snippets.

@danieltomasz
Created January 29, 2023 23:07
Show Gist options
  • Save danieltomasz/5964de9c15ef918ad505d8f43c1131de to your computer and use it in GitHub Desktop.
Save danieltomasz/5964de9c15ef918ad505d8f43c1131de to your computer and use it in GitHub Desktop.
Iterating over variables in glm
create_models <- function(datame_name, cog_vars, eeg_components) {
list_models <- list()
for (cogvar in cog_vars) { # ,eeg_components
for (eegvar in eeg_components) { # ,eeg_components
models <- glue("model1.{cogvar}.{eegvar} <- glm({cogvar} ~ GM_Normalised + {eegvar}, {datame_name}, family = gaussian)
list_models[['model.{cogvar}.{eegvar}']] <- model5.{cogvar}.{eegvar}") # nolint
eval(str2expression(models))
}
}
return(list_models)
}
list_models <- create_models("dataset", cog_vars, eeg_components)
df_models <- purrr::map_df(list_models, broom::tidy, .id = "model")
print.data.frame(df_models, scientific = FALSE, digits = 2)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment