r_eda

dot 和 bar 的比较

参考 Strayer (2019)

当表示大小时，点图是要比 bar 更加清晰。

knitr::opts_chunk$set(warning = FALSE, message = FALSE)
suppressMessages(source(here::here("R/load.R")))

who_disease <- read_csv('datasets/who_disease.csv')
interestingCountries <- c("NGA","SDN","FRA","NPL","MYS","TZA","YEM","UKR","BGD","VNM")
who_subset <- who_disease %>% 
    filter(
        countryCode %in% interestingCountries,
        disease == 'measles',
        year %in% c(1992, 2002) # Modify years to 1992 and 2002
    ) %>% 
    mutate(year = paste0('cases_', year)) %>% 
    spread(year, cases)
 
# Reorder y axis and change the cases year to 1992
ggplot(who_subset, aes(x = log10(cases_1992), y = reorder(country,cases_1992))) +
    geom_point()

如图是查看若干国家，在 1992年疾病案件的数量比较。 但是这个图是有 track 的，感觉 Nepal 很少，我们不妨查看增长率。

who_subset %>% 
    # calculate the log fold change between 2016 and 2006
    mutate(logFoldChange = log2(cases_2002/cases_1992)) %>% 
    # set y axis as country ordered with respect to logFoldChange
    ggplot(aes(x = logFoldChange, y = reorder(country,logFoldChange))) +
    geom_point() +
    # add a visual anchor at x = 0
    geom_vline(xintercept = 0)

在这里会发现，

1. Nepal 虽然基数最小，但是增长最快
2. France 基数最大，但是增长最慢 (且是下降)

geom_vline(xintercept = 0) 给了一个 reference line，表示1992年和2006两年案件数量没有增长和下降。

这里可以总结一下， bar 图在这里的表现就会很差，

1. 使用 log 会很奇怪
2. reference line 也有没 dot 图清晰

下面加入地区

who_subset %>% 
    mutate(logFoldChange = log2(cases_2002/cases_1992)) %>% 
    ggplot(aes(x = logFoldChange, y = reorder(country, logFoldChange))) +
    geom_point() +
    geom_vline(xintercept = 0) +
    xlim(-6,6) +
    # add facet_grid arranged in the column direction by region and free_y scales
    facet_grid(region ~ ., scales = 'free_y')

如果使用 bar，那么图片就很臃肿。

Cleaning up the bars

amr_pertussis <- who_disease %>% 
    filter(   # filter data to our desired subset
        region == 'AMR', 
        year == 1980, 
        disease == 'pertussis'
    )
# Set x axis as country ordered with respect to cases. 
ggplot(amr_pertussis, aes(x = reorder(country,cases), y = cases)) +
    geom_col() +
    # flip axes
    coord_flip()

1. 第一个问题是，有很多 0 cases 在，可以剔除。
2. 横线可以剔除，比较碍事

amr_pertussis %>% 
    # filter to countries that had > 0 cases. 
    filter(cases > 0) %>%
    ggplot(aes(x = reorder(country, cases), y = cases)) +
    geom_col() +
    coord_flip() +
    theme(
    # get rid of the 'major' y grid lines
    panel.grid.major.y = element_blank()
    )

最好的方式是换成点图

amr_pertussis %>% filter(cases > 0) %>% 
    ggplot(aes(x = reorder(country, cases), y = cases)) + 
    # switch geometry to points and set point size = 2
    geom_point(size = 2) + 
    geom_segment( aes(x=reorder(country, cases), xend= reorder(country, cases), y=0, yend= cases)) +
    # change y-axis to log10. 
    scale_y_log10() +
    # add theme_minimal()
    theme_minimal() +
    coord_flip()

然后转换成棒棒糖图，参考 Reference

参考 WeChat Article，增加面积图使得图像更形象。

geom_point type

amr_pertussis %>% filter(cases > 0) %>% 
    mutate(country = country %>% as.factor %>% fct_reorder(cases),
           country_index = country %>% as.integer) %.>% 
    ggplot(data = .) + 
    aes(x = country_index, y = cases) +
    geom_area(color = 'black', fill = 'black', alpha = .2) +
    geom_segment(aes(xend= country_index, 
                     y=0, yend= cases),
                 colour="grey50", linetype="dashed",
                 alpha = .2
                 ) +
    geom_point(size = 2, shape = 21, fill = 'white', col = 'black') + 
    # vignette("ggplot2-specs") 
    # 查看 shape
    # 注意 geom_point 放在 geom_area 之后，才能够保证点中空
    scale_x_continuous(breaks = .$country_index, labels = .$country) +
    geom_text(size = 2, aes(label = cases), nudge_y = 0.2) + 
    scale_y_log10() +
    # add theme_minimal()
    theme_minimal() +
    theme(
       panel.grid = element_blank()
    ) +
    coord_flip()

Strayer, Nick. 2019. “Visualization Best Practices in R.” DataCamp. 2019. <https://www.datacamp.com/courses/visualization-best-practices-in-r>.