# tested only for family poisson for now...
glm_out <- function(fit_model = fit_model){
  require(tidyverse)
  data <- fit_model[["model"]]
  # define crit betas as Belsley et al. 1980
  critbetas <- 2/sqrt(NROW(data))
  betas <- dfbeta(fit_model) %>% 
    as_tibble() %>% select(-`(Intercept)`) %>% 
    # changing all to absolute values
    mutate(across(everything(), ~abs(.x))) %>% 
    # possible outliers will be coded as 1
    mutate(across(everything(), ~ifelse(.x > critbetas,1,0))) %>% 
    mutate(total = rowSums(across(where(is.numeric)))) %>% 
    mutate(total = ifelse(total >= 1, 1,0)) %>% 
    select(total)
  pbeta <- dfbeta(fit_model) %>% 
    as_tibble() %>% select(-`(Intercept)`) %>% 
    rownames_to_column() %>% 
    mutate(rowname = as.numeric(rowname)) %>% 
    rename(index = rowname) %>% 
    pivot_longer(cols = c(-index)) %>% 
    ggplot(aes(x = index, y = value)) + 
    geom_point()+
    geom_jitter()+
    facet_wrap(name~.)+
    geom_hline(yintercept = critbetas, color = "red")+
    geom_hline(yintercept = 0-critbetas, color = "red")+
    theme_classic()+
    theme(axis.text.x = element_text(colour="black",size=10),
          panel.background = element_blank(), 
          axis.line = element_line(colour = "black"),
          axis.text.y = element_text(colour="black",size=11))+
    labs(title = "DF betas", 
         x = "case",y = "b value")
  
  # define crit cooks as Beaujean et al. 2016
  cooks<-data %>% 
    mutate(cooks_dist = cooks.distance(fit_model)) %>% 
    mutate(out_cook = ifelse(cooks_dist>(4/NROW(.)),1,0))
  noout_data <- cbind(cooks, betas) %>% as_tibble() %>% 
    mutate(total_out = out_cook + total) %>% 
    filter(total_out < 2) %>% # exclude people who met the 2 criteria
    select(-total_out, -out_cook, -total, -cooks_dist)
  # user messages
  message_dfbetas = paste(c("the cut off for DFBETAS values was ",
                            critbetas), collapse = '')
  
  message_cooks = paste(c("the cut off for cooks values was ",
                          4/NROW(data)), collapse = '')
  # outliers excluded
  print(paste(c("a total of ", (NROW(data) - NROW(noout_data)), 
                " outliers should be excluded, i.e., cases that met the two criteria"), collapse = ''))
  out_betas <- betas %>% count(total)
  out_cooks <- cooks %>% count(out_cook)
  output = list(message_dfbetas = message_dfbetas,
                message_cooks = message_cooks,
                out_betas = out_betas, 
                out_cooks = out_cooks, 
                noout_data = noout_data,
                plot_betas = pbeta)
  return(output)
}