EXPLORATORY DATA ANALYSIS. Introducing the data

Transcription

1 EXPLORATORY DATA ANALYSIS Introducing the data

2 data set > # A tibble: 3, spam to_multiple from cc sent_ time image <fctr> <dbl> <dbl> <int> <dbl> <dttm> <dbl> 1 not-spam :16: not-spam :03: not-spam :00: not-spam :09: not-spam :00: not-spam :04: not-spam :55: not-spam :45: not-spam :08: not-spam :12:00 0 #... with 3,911 more rows, and 14 more variables: attach <dbl>, # dollar <dbl>, winner <fctr>, inherit <dbl>, viagra <dbl>, # password <dbl>, num_char <dbl>, line_breaks <int>, format <dbl>, # re_subj <dbl>, exclaim_subj <dbl>, urgent_subj <dbl>, # exclaim_mess <dbl>, number <fctr>

3 Histograms > ggplot(data, aes(x = var1)) + geom_histogram()

4 Histograms > ggplot(data, aes(x = var1)) + geom_histogram() + facet_wrap(~var2)

5 Boxplots > ggplot(data, aes(x = var2, y = var1)) + geom_boxplot()

6 Boxplots > ggplot(data, aes(x = 1, y = var1)) + geom_boxplot()

7 Density plots > ggplot(data, aes(x = var1)) + geom_boxplot()

8 Density plots > ggplot(data, aes(x = var1, fill = var2)) + geom_density(alpha =.3)

9 EXPLORATORY DATA ANALYSIS Let s practice!

10 EXPLORATORY DATA ANALYSIS Check-in 1

11 Review

12 Zero inflation strategies Analyze the two components separately Collapse into two-level categorical variable

13 Zero inflation strategies > %>% mutate(zero = exclaim_mess == 0) %>% ggplot(aes(x = zero)) + geom_bar() + facet_wrap(~spam)

14 Barchart options > %>% mutate(zero = exclaim_mess == 0) %>% ggplot(aes(x = zero, fill = spam)) + geom_bar()

15 Barchart options > %>% mutate(zero = exclaim_mess == 0) %>% ggplot(aes(x = zero, fill = spam)) + geom_bar(position = "fill")

16 EXPLORATORY DATA ANALYSIS Let s practice!

17 EXPLORATORY DATA ANALYSIS Check-in 2

18 Spam and images > %>% mutate(has_image = image > 0) %>% ggplot(aes(x = as.factor(has_image), fill = spam)) + geom_bar(position = "fill")

19 Spam and images > %>% mutate(has_image = image > 0) %>% ggplot(aes(x = spam, fill = has_image)) + geom_bar(position = "fill")

20 Ordering bars

21 Ordering bars > <- %>% mutate(zero = exclaim_mess == 0) > levels( $zero) NULL > $zero <- factor( $zero, levels = c("true", "FALSE")) > %>% ggplot(aes(x = zero)) + geom_bar() + facet_wrap(~spam) TRUE first, then FALSE

22 Ordering bars > <- %>% mutate(zero = exclaim_mess == 0) > levels( $zero) NULL > $zero <- factor( $zero, levels = c("true", "FALSE")) > %>% ggplot(aes(x = zero)) + geom_bar() + facet_wrap(~spam)

23 EXPLORATORY DATA ANALYSIS Let s practice!

24 EXPLORATORY DATA ANALYSIS Conclusion

25 Pie chart vs. bar chart

26 Faceting vs. stacking

27 Histogram > ggplot(data, aes(x = var1)) + geom_histogram()

28 Density plot > cars %>% filter(eng_size < 2.0) %>% ggplot(aes(x = hwy_mpg)) + geom_density()

29 Side-by-side box plots > ggplot(common_cyl, aes(x = as.factor(ncyl), y = city_mpg)) + geom_boxplot() Warning message: Removed 11 rows containing non-finite values (stat_boxplot).

30 Center: mean, median, mode > x [1] > table(x) x

31 Shape of income > ggplot(life, aes(x = income, fill = west_coast)) + geom_density(alpha =.3) > ggplot(life, aes(x = log(income), fill = west_coast)) + geom_density(alpha =.3)

32 With group_by() > life %>% + slice(240:247) %>% + group_by(west_coast) %>% + summarize(mean(expectancy)) # A tibble: 2 x 2 west_coast mean(expectancy) <lgl> <dbl> 1 FALSE TRUE state county expectancy income west_coast California Tuolumne TRUE California Ventura TRUE California Yolo TRUE California Yuba TRUE Colorado Adams FALSE Colorado Alamosa FALSE Colorado Arapahoe FALSE Colorado Archuleta FALSE

33 Spam and exclamation points > %>% mutate(zero = exclaim_mess == 0) %>% ggplot(aes(x = zero, fill = spam)) + geom_bar()

34 Spam and images > %>% mutate(has_image = image > 0) %>% ggplot(aes(x = as.factor(has_image), fill = spam)) + geom_bar(position = "fill")

35 EXPLORATORY DATA ANALYSIS Congratulations!