tuna: See what may have passed through your glmnet.

summarize_repeated_lasso <- function(x){
  df = x[["df"]]
  df %>% dplyr::group_by(var) %>%
    dplyr::select(n, mean, sd, descr) %>%
    dplyr::slice(1)
}


#' visualize_all_cverror
#' 
#' shows all CV error for all lambdas across all intialization conditoins
#'
#' @param x list generated by the function tuna::repeat_lasso
#'
#' @return plots in r base graphics
#' @export
visualize_all_cverror <- function(x){
  plot(log10(x[["fits"]][[1]]$lambda), x[["fits"]][[1]]$cvm, type = "n",
       ylab = x[["fits"]][[1]]$name,
       xlab = expression(log[10]~"(lambda)"))

  mm <- x[["fits"]][[1]]$nzero

  axis(3, at = log10(x[["fits"]][[1]]$lambda), labels = mm, cex.axis= .5)
  purrr::walk(x[["fits"]], ~points(log10(.x$lambda), .x$cvm,type = "l", col = '#00000020'))
  purrr::walk(x[["fits"]], ~abline(v = log10(.x$lambda.min), lty = 1, col = '#0000FF20'))


  purrr::map(x[["fits"]], ~points(log10(x = .x$lambda.min),
                                     y = .x$cvm[which(.x$lambda == .x$lambda.min)],
                                     cex = .8, pch = 20, col = '#0000FF20'))



  cvms <- purrr::map(x[["fits"]], function(x) x$cvm)
  cvm_mean <- apply(purrr::invoke(cbind, cvms),1, mean)

  points(log10(x[["fits"]][[1]]$lambda), cvm_mean, col = "black", type = "l", lty =2,  lwd = 2)
}


#' visualize_coef_z_density
#' 
#' density curves (z-scores) for standardized coefficient values
#'
#' @param x list generated by the function tuna::repeat_lasso
#'
#' @return ggplot2 object
#' @export
visualize_coef_z_density = function(x){
  df = x[["df"]]
  ggplot(df, aes(x = z)) +
    geom_density(fill = "black") +
    theme_minimal() +
    theme(axis.text.x = element_text(angle = 90, size = 6),
          axis.text.y = element_text( size = 8),
          strip.text.x = element_text(size = 8, colour = "red")) +
    facet_wrap(~descr, scale = "free_y") +
    ylab("") +
    xlab("coef Z")
}


#' visualize_coef_density
#' 
#' density curves for raw coefficient values
#'
#' @param x list generated by the function tuna::repeat_lasso
#'
#' @return ggplot2 object
#' @export
visualize_coef_density = function(x){
  df = x[["df"]]
  df <- df %>% dplyr::mutate(sign = sign(mean)) %>%
    dplyr::mutate(cl = ifelse(sign > 0, "pos", "neg"))

  ggplot(df, aes(x = coef, fill = cl, col = cl)) +
    geom_density() +
    theme_minimal() +
    theme(axis.text.x = element_text(angle = 90, size = 6),
          axis.text.y = element_text( size = 8),
          strip.text.x = element_text(size = 8, colour = "red")) +
    theme(legend.position = "none") +
    facet_wrap(~descr, scale = "free_y") +
    scale_fill_manual(values = c("#17becf", "#bcbd22")) +
    scale_color_manual(values = c("#17becf", "#bcbd22")) +
    ylab("Density") +
    xlab("Coefficient (Z-score)") +
    geom_vline(xintercept = 0, lty = 2, lwd = .5)
}

visualize_coef_histograms <- function(x){
  df = x[["df"]]
  ggplot(df, aes(x = coef)) +
    geom_histogram(fill = "black") +
    theme_minimal() +
    theme(axis.text.x = element_text(angle = 90, size = 6),
          axis.text.y = element_text( size = 8),
          strip.text.x = element_text(size = 8, colour = "red")) +
    facet_wrap(~descr, scale = "free") +
    xlab("") +
    ylab("")
}

get_lambda_mins <- function(x){
  lambda_mins <- purrr::map(x[["fits"]], ~return(.x$lambda.min))
}

get_cv_mins <- function(x){
  cv_mins <- purrr::map(x[["fits"]], ~return(.x$cvm[which(.x$lambda == .x$lambda.min)]))
}
get_coefs_df <- function(x){
  coefs <- purrr::map(x[["fits"]], ~return(as.matrix(coef(.x, lambda = .x$lambda.min))))
  coefs_df<- purrr::map_df(coefs, ~return(data.frame(var = rownames(.x), coef = .x[,1])))
}