פרק 8 - ggplot

החבילה ggplot2 מסייעת להציג מידע בצורה גרפית

# install the library ggplot2 if it is not already installed
if (!require("ggplot2")) install.packages("ggplot2")
# load the library ggplot2
library("ggplot2")

הורד את הקובץ songs.csv

---

דוגמא 1: QQplot לצורך בדיקת הנחת הנורמליות בשתי קבוצות (שני גרפים זה לצד זה)

if (!require("ggplot2")) install.packages('ggplot2') # for data visualization
if (!require("tidyverse")) install.packages('tidyverse') # dealing with dataframes
if (!require("tidylog")) install.packages('tidylog') # logs for tidyverse
if (!require("plotrix")) install.packages('plotrix') # for the standard error function

songsDataset <- read.csv('songs.csv') # read the dataset CSV file 
songsDataset <- na.omit(songsDataset) # to remove NA values if there are (makes no change)

# new categorical column "explicit_text" to translate the binary column "explicit"
songsDataset <- songsDataset %>% 
  mutate(explicit_text = case_when(
    explicit == 0 ~ "Implicit",
    explicit == 1 ~ "Explicit",
  ))

# summary table that holds descriptive statistics about the variable popularity:
# mean, sample size, standard deviation, standard error of the sample and median
# for both groups - Explicit songs and Implicit songs
stats_popularity_per_type <- songsDataset %>% 
  group_by(explicit_text) %>% 
  summarise(popularity_mean = mean(popularity), 
            n=n(),
            std = sd(popularity),
            sterr = std.error(popularity),
            median = median(popularity))

# simultanious qqplot for the variable popularity in both Implicit & Explicit songs
ggplot(songsDataset) + 
  geom_qq(aes(sample = popularity, color=explicit_text), size=1) +
  geom_qq_line(aes(sample = popularity)) +
  facet_wrap(~explicit_text, ncol = 6, shrink = TRUE) +
  guides(color='none') +
  labs(x='Theoretical Z score', y='Popularity',
       title = 'QQplot for the variable popularity in both Implicit & Explicit songs',
       subtitle = 'To check the noramality assumption on each group') +
  theme(plot.title = element_text(color="grey20",size=14, face="bold.italic"),
    plot.subtitle = element_text(color="grey20", face="italic"))

---

דוגמא 2: הצגה ויזואלית של מבחן t למדגמים בלתי תלויים + סטטיסטיקה תיאורית

ggplot(songsDataset, aes(x=popularity, fill = explicit_text)) +
  geom_density(alpha=0.5) +
  geom_vline(data = stats_popularity_per_type, 
             aes(xintercept = popularity_mean), linetype="dashed") +
  geom_text(data = stats_popularity_per_type, 
            aes(x = 87.5, y = 0.03, label = paste('N:', n), 
                color = explicit_text), 
            size = 4) + 
  geom_text(data = stats_popularity_per_type, 
            aes(x = 87.5, y = 0.025, label = paste('Mean:',round(popularity_mean,2)), 
                color = explicit_text), 
            size = 4) +
  geom_text(data = stats_popularity_per_type, 
            aes(x = 87.5, y = 0.02, label = paste('Median:',round(median,2)), 
                color = explicit_text), 
            size = 4) +
  geom_text(data = stats_popularity_per_type, 
            aes(x = 87.5, y = 0.015, label = paste('Std:',round(std,2)), 
                color = explicit_text), 
            size = 4) +  
  geom_text(data = stats_popularity_per_type, 
            aes(x = 87.5, y = 0.01, label = paste('Sterr:',round(sterr,2)), 
                color = explicit_text), 
            size = 4) + 
  facet_wrap(~explicit_text, ncol=1) +
  guides(color='none', fill='none') +
  scale_fill_manual(values = c('cornflowerblue', 'darkgoldenrod')) +
  scale_color_manual(values = c('cornflowerblue', 'darkgoldenrod')) +
  labs(title = 'Is there a difference in popularity between Explicit and Implicit songs?',
       subtitle = 'each vertical line represents the sample mean',
       y='Density',x='Popularity grade') +
  theme(plot.title = element_text(color="grey25",size=13.5, face="bold.italic"),
        plot.subtitle = element_text(color="grey25", face="italic"))

---

דוגמא 3: QQplot לצורך בדיקת הנחת הנורמליות באחת עשרה קבוצות (11 גרפים זה לצד זה)

# extract the year, century and decade
songsDataset = songsDataset %>%
  # extract only the 4 digits represents the year - 1996, 1956, 1977 etc
  mutate(release_year = str_extract_all(release_date, "\\d{4}")) %>% 
  # 1900s, 2000s
  mutate(release_century = paste(substr(release_year,1,2), '00s', sep='')) %>%
  # 1920s, 1950s, 2010s, etc
  mutate(release_decade = paste(substr(release_year,1,3), '0s', sep='')) %>% 
  # make sure the data is OK. songs should be from 1922 to 2021
  filter(release_year >= 1922 & release_year <= 2021) 

# summary table that holds descriptive statistics about the variable tempo:
# mean, sample size, standard deviation, standard error of the sample and median
# for all decades - from 1920s to 2020s
stats_relase_decade_and_tempo = songsDataset %>%
  group_by(release_decade) %>%
  summarise(tempo_mean = round(mean(tempo),2), 
            n=n(), 
            std = round(sd(tempo),2),
            sterr = round(std.error(tempo),3),
            median = round(median(tempo),2)) %>% 
  # order the stats table by the decade
  arrange(release_decade) 
  
# simultaneous qqplot for the variable tempo in all decades
ggplot(songsDataset) + # it takes some time to run... a lot of data to process
  geom_qq(aes(sample = tempo, color=release_decade), size=0.5) +
  geom_qq_line(aes(sample = tempo)) +
  facet_wrap(~release_decade, ncol = 6, shrink = TRUE) +
  guides(color='none') +
  labs(x='Theoretical Z score', y='Tempo',
       title = 'QQplot for the variable tempo in all decades',
     subtitle = 'To check the noramality assumption on each group') +
  theme(#aspect.ratio=1,
        plot.title = element_text(color="grey20",size=14, face="bold.italic"),
        plot.subtitle = element_text(color="grey20", face="italic"))

---

דוגמא 4: הצגה ויזואלית של ניתוח שונות חד גורמי עם 11 קבוצות + סטטיסטיקה תיאורית

ggplot(songsDataset, aes(x=tempo, fill=release_decade)) +
    geom_density(alpha = 0.5) +
    facet_wrap(~release_decade, ncol = 3,shrink = TRUE, dir="v") +
    geom_vline(data = stats_relase_decade_and_tempo, 
               aes(xintercept = tempo_mean), linetype="dashed") +
    geom_text(data = stats_relase_decade_and_tempo, 
              aes(x = 200, y = 0.015, label = paste('Mean:',round(tempo_mean,2)), 
                  color = release_decade), size = 2.8) +
    geom_text(data = stats_relase_decade_and_tempo, 
              aes(x = 200, y = 0.011, label = paste('Median:',round(median,2)), 
                  color = release_decade), size = 2.8) +
    geom_text(data = stats_relase_decade_and_tempo, 
              aes(x = 200, y = 0.007, label = paste('sd:',round(std,2)), 
                  color = release_decade), size = 2.8) +
    labs(x = 'Tempo', y='Density',
         title = 'Density distribuition of Tempo by Decade',
        subtitle = 'each vertical line represents the sample mean') +
    guides(color='none', fill='none') +
    theme(plot.title = element_text(color="grey20",size=14, face="bold.italic"),
          plot.subtitle = element_text(color="grey20", face="italic"))

---