###########################
# Nowe dane
###########################

load("diagnozaOS.Rdata")
#attach(dane)
#head(dane)

#dane[, "bp124"]
#dane <- dane[which(! is.na(dane[, "bp124"])), ]

# Wczytanie danych
readData <- function() {
  dane <<- dane[which(! is.na(dane[, "bp124"])), ]
  dane <- na.omit(dane)
#   for (i in colnames(dane)) {
#     dane <<- dane[, which(length(levels(dane[, i])) > 1)]
#   }
}

#readData()

lB = 0.05
uB = 0.95

# Obcięcie skrajnych wartosci
trimData <- function(lowerBound = lB, upperBound = uB) {
  dane <<- dane[dane[, "bp124"] > quantile(dane[, "bp124"], probs = lowerBound), ]
  dane <<- dane[dane[, "bp124"] < quantile(dane[, "bp124"], probs = upperBound), ]
}

readData()
trimData(lowerBound = lB, upperBound = uB)

banned.cols <- c(44, 216:242, 345, 355:364, 626, 896:920, 1531:1532, 1825, 1859, 2014:2021)
banned.cols <- c(banned.cols, 93:115)
# 896:920 -- troche za wyrost


find.valuable.cols <- function(p.val.max = 0.05) {

  full.form <- "bp124 ~ "
  dane.cols.num <- length(colnames(dane))

  result.i = numeric(50)
  p.val = numeric(50)
  licznik <- 0
  for (i in colnames(dane)) {
    dane.tmp <- dane[, c("bp124", i)]
#     if (nlevels(dane[, i]) > 1 & ! is.element(which(colnames(dane) == i), banned.cols)) {
    if (length(levels(dane[, i])) > 1 & ! is.element(which(colnames(dane) == i), banned.cols)) {
      if (! i == colnames(dane)[1]) full.form <- paste(full.form, "+")
      full.form <- paste(full.form, i)

      fmla <- as.formula(paste("bp124", "~", colnames(dane.tmp)[2], sep = ''))
      model.tmp <- lm(formula = fmla, data = dane.tmp)

      if (! is.na(anova(model.tmp)[1, "Pr(>F)"]) & (anova(model.tmp)[1, "Pr(>F)"] <= p.val.max)) {
#         print(anova(model.tmp))
#         print(i)
        result.i[licznik] <- i
        p.val[licznik] <- anova(model.tmp)[1, "Pr(>F)"] # p.val.tmp
        licznik <- licznik + 1
      }
    }
  }

  list(i = result.i, pv = p.val, l = licznik, f = full.form)
}

lst1 <- find.valuable.cols(p.val.max = 0.0000000000000000000000000000005) # 15
lst2 <- find.valuable.cols(p.val.max = 0.00000000000000000000000005) # 10
lst1$i
lst2$i