# Balíčky
library(psych)
library(haven)
library(tidyverse)

df <- read_sav("data/hw_inventory.sav")
meta <- read_rds("data/hw_metadata.rds")

glimpse(df)
names(df)



# Tvorba nových proměnných (sloupců) pomocí mutate() --------------------------
df <- mutate(df,
             sr_height_m = sr_height_cm / 100,
             bmi = sr_weight_kg / sr_height_m^2)

select(df, sr_height_m, bmi)

# Alternativou k funkci mutate() je funkce transmute()
transmute(df,
          sr_height_m = sr_height_cm / 100,
          bmi = sr_weight_kg / sr_height_m^2) 

# Pokud chceme změny uložit
df <- df %>% 
  mutate(bmi = sr_weight_kg / (sr_height_cm/100)^2)


# Převedení vybraných proměnných na proměnné kategorické (factors)
summary(df$gender)
summary(df[["gender"]])

summary(df$sr_wcat)

df <- df %>% 
  mutate(gender = as_factor(gender),
         sr_wcat = as_factor(sr_wcat))

summary(df$gender)
summary(df$sr_wcat)

# List položek
c("hi_01", "hi_02")

str_c("hi_", 1:26)

str_pad(1:26, width = 2, pad = 0)

str_c("hi_", 
      str_pad(1:26, width = 2, pad = 0))

str_c("wi_", 
      str_pad(1:22, width = 2, pad = 0))

str_c("rses_", 
      str_pad(1:10, width = 2, pad = 0))

items <- list(hi = str_c("hi_", 
                         str_pad(1:26, width = 2, pad = 0)),
              wi = str_c("wi_",
                         str_pad(1:22, width = 2, pad = 0)),
              rses = str_c("rses_",
                           str_pad(1:10, width = 2, pad = 0)))
items

# Reverzní položky
rev_items <- c(str_c("hi_", 
                     str_pad(14:26, width = 2, pad = 0)),
               str_c("wi_",
                     str_pad(12:22, width = 2, pad = 0)),
               str_c("rses_",
                     str_pad(c(2, 5, 6, 8, 9), width = 2, pad = 0)))

# Korelace mezi položkami
df %>% 
  select(all_of(items$hi)) %>% 
  cor(use = "pairwise") %>% 
  round(digits = 2)

# Rekódování reverzních položek
df %>% 
  mutate(hi_14 = 5 - hi_14,
         hi_15 = 5 - hi_15)

?across

df %>% 
  transmute(across(hi_14:hi_26, ~5 - .x))

df %>% 
  transmute(across(hi_14:hi_26, ~5 - .x,
                   .names = "{.col}_rec"))

df_old <- df

df <- df %>% 
  mutate(across(all_of(rev_items), ~5 - .x))
 
# Položková analýza
df %>% 
  select(all_of(items$hi)) %>% 
  psych::alpha()

df %>% 
  select(all_of(items$hi)) %>% 
  psych::omega(nfactors = 2)

# Složitější rekódování
  # Below 18.5	Underweight
  # 18.5 – 24.99	Healthy Weight
  # 25.0 – 29.99	Overweight
  # 30.0 and Above	Obesity

summary(df$bmi)

bmi_class <- c("Underweight", "Healthy Weight", "Overweight",
               "Obesity")

df %>% 
  filter(!between(bmi, 15, 40)) %>% 
  select(id, gender, age, sr_height_cm, sr_weight_kg, bmi) %>% 
  arrange(bmi)

df %>% 
  mutate(bmi_cat = case_when(bmi < 18.5 ~ bmi_class[1],
                             bmi < 25 ~ bmi_class[2],
                             bmi < 30 ~ bmi_class[3],
                             bmi >= 30 ~ bmi_class[4],
                             TRUE ~ NA_character_) %>% 
           factor(levels = bmi_class)) %>% 
  select(bmi, bmi_cat) %>% 
  print(n = 30)
  
df <- df %>% 
  mutate(bmi_cat = case_when(bmi < 18.5 ~ bmi_class[1],
                             bmi < 25 ~ bmi_class[2],
                             bmi < 30 ~ bmi_class[3],
                             bmi >= 30 ~ bmi_class[4],
                             TRUE ~ NA_character_) %>% 
           factor(levels = bmi_class))


df %>% 
  transmute(across(all_of(items$hi), ~case_when(.x <= 2 ~ 0L,
                                                .x >= 3 ~ 1L,
                                                TRUE ~ NA_integer_)))

# Suspicious cases
df_old %>% 
  filter(if_all(hi_01:hi_26, ~.x == 4))

df_old %>% 
  filter(if_all(hi_01:hi_26, ~.x == 1))

df <- df %>% 
  filter(id != 2184)

# computing total scores
df %>% 
  transmute(total_hi = hi_01 + hi_02 + hi_03)

df %>% 
  transmute(total_hi = pick(all_of(items$hi)) %>% 
              rowMeans())

df %>% 
  transmute(total_hi = pick(all_of(items$hi)) %>% 
              rowMeans(na.rm = TRUE))

df %>% 
  transmute(total_hi = pick(all_of(items$hi)) %>% 
              rowMeans(na.rm = TRUE),
            na_hi = pick(all_of(items$hi)) %>% 
              is.na() %>% 
              rowSums(),
            total_hi = if_else(na_hi > 1, NA_real_, total_hi))

row_mean <- function(..., max_na = 0) {
  
  # Vybrat sloupce, které chceme
  data <- pick(...)
  
  # Vypočíst součet chybějícíh dat (počet NA) pro daný řádek
  n_missing <- data %>% 
    is.na() %>% 
    rowSums()
  
  # Vypočíst celkový skór jako průměr položek pro daný řádek
  out <- data %>% 
    rowMeans(na.rm = TRUE)
  
  # Nahradit celkový skór chybějící hodnotou, pokud je n_missing > max_na
  out[n_missing > max_na] <- NA_real_
  
  # Výstupní hodnota (output funkce)
  return(out) 
}

df %>% 
  transmute(total_hi = row_mean(all_of(items$hi), max_na = 0))

df %>% 
  transmute(total_hi = row_mean(all_of(items$hi), max_na = 1))

df <- df %>% 
  mutate(total_hi = row_mean(all_of(items$hi), max_na = 2))


# Souhrnné statistiky pomocí summarise() ------------------------------------
df %>% 
  summarise(M_height = mean(sr_height_cm))

df %>% 
  summarise(M_height = mean(sr_height_cm, na.rm = TRUE))


# To, jak si nově vypočtenou statistiku pojmenujeme, je na nás.
df %>% 
  summarise(průměrná_výška = mean(sr_height_cm, na.rm = TRUE))

# Užitečnější je ovšem tato funkce v kombinaci s funkcí group_by()
by_gender <- group_by(df, gender)
by_gender

by_gender %>% 
  summarise(M_height = mean(sr_height_cm, na.rm = TRUE))

# Protože často potřebujeme dataset rozdělit různě a vytvářet pokaždé nový, 
# jinak rozdělený dataset (jako v tomto případě by_day) je poněku zbytečné, 
# často využíváme pipe (%>%) ke zřetězení více operací dohromady.

df %>% # použitý dataset
  group_by(gender, sr_wcat) %>% # rozdělit podle
  summarise(M_bmi = mean(bmi, na.rm = TRUE)) %>% # Vypočíst statistiky
  arrange(sr_wcat, gender) # seřadit podle

df %>%
  group_by(gender, sr_wcat) %>% 
  summarise(M_bmi = mean(bmi, na.rm = TRUE),
            SD_bmi = sd(bmi, na.rm = TRUE),
            n = n()) %>%
  arrange(sr_wcat, gender)


df %>%
  drop_na(gender, sr_wcat, bmi) %>% 
  group_by(gender, sr_wcat) %>% 
  summarise(M_bmi = mean(bmi, na.rm = TRUE),
            SD_bmi = sd(bmi, na.rm = TRUE),
            n = n()) %>%
  arrange(sr_wcat, gender)

# Více proměnných
df %>% 
  group_by(gender) %>% 
  summarise(across(c(sr_height_cm, sr_weight_kg, bmi), 
                   ~mean(.x, na.rm = TRUE),
                   .names = "M_{.col}"))

df %>% 
  group_by(gender) %>% 
  summarise(across(c(sr_height_cm, sr_weight_kg, bmi), 
                   ~mean(.x, na.rm = TRUE),
                   .names = "M_{.col}"))

df %>% 
  select(gender, sr_height_cm, sr_weight_kg, bmi)

df %>% 
  select(id, gender, sr_height_cm, sr_weight_kg, bmi)

df %>% 
  select(id, gender, sr_height_cm, sr_weight_kg, bmi) %>% 
  pivot_longer(cols = c(sr_height_cm, sr_weight_kg, bmi),
               names_to = "variable", values_to = "value")

df %>% 
  select(id, gender, sr_height_cm, sr_weight_kg, bmi) %>% 
  pivot_longer(cols = c(sr_height_cm, sr_weight_kg, bmi),
               names_to = "variable", values_to = "value") %>% 
  drop_na(gender, value)

df %>% 
  select(id, gender, sr_height_cm, sr_weight_kg, bmi) %>% 
  pivot_longer(cols = c(sr_height_cm, sr_weight_kg, bmi),
               names_to = "variable", values_to = "value") %>% 
  drop_na(gender, value) %>% 
  group_by(gender, variable) %>% 
  summarise(M = mean(value),
            SD = sd(value),
            Min = min(value),
            q1 = quantile(value, probs = .25),
            Mdn = median(value),
            q3 = quantile(value, probs = .75),
            Max = max(value),
            skew = psych::skew(value),
            kurt = psych::kurtosi(value))

df %>% 
  drop_na(gender, sr_wcat) %>% 
  count(gender, sr_wcat)

df %>% 
  drop_na(gender, sr_wcat) %>% 
  count(gender, sr_wcat) %>% 
  group_by(gender) %>% 
  mutate(p = n / sum(n))