gapminder dataset
This dataset (available in eponymous package on CRAN) provides values for life expectancy, GDP per capita, and population, every five years, from 1952 to 2007, for each of 142 countries and was collected by the Gapminder Foundation. Let’s have a look at the data-
library(gapminder)
library(dplyr)
library(magrittr)
gapminder::gapminder %>%
dplyr::glimpse(x = .)
#> Observations: 1,704
#> Variables: 6
#> $ country <fct> Afghanistan, Afghanistan, Afghanistan, Afghanistan, ...
#> $ continent <fct> Asia, Asia, Asia, Asia, Asia, Asia, Asia, Asia, Asia...
#> $ year <int> 1952, 1957, 1962, 1967, 1972, 1977, 1982, 1987, 1992...
#> $ lifeExp <dbl> 28.801, 30.332, 31.997, 34.020, 36.088, 38.438, 39.8...
#> $ pop <int> 8425333, 9240934, 10267083, 11537966, 13079460, 1488...
#> $ gdpPercap <dbl> 779.4453, 820.8530, 853.1007, 836.1971, 739.9811, 78...
Distribution with gghistostats
Suppose the first thing we want to check is the distribution of population worldwide in 2007. In this case, we are not interested in any statistics and, therefore, can set the results.subtitle argument to FALSE.
Although this plot is useful, it is still not satisfactory as most of the mass seems to be concentrated at 0 due to the large range of numbers. We can remedy this by converting population to logarithmic scale. We can additionally adjust binwidth so that we have bins for every increase in order of magnitude.
gapminder::gapminder %>%
dplyr::filter(.data = ., year == 2007) %>% # select data only from the year 2007
dplyr::mutate(.data = ., pop_log = log10(pop)) %>% # creating new population variable
ggstatsplot::gghistostats(
data = ., # data from which variable is to be taken
x = pop_log, # numeric variable
results.subtitle = FALSE, # don't run statistical tests
messages = FALSE, # turn off messages
xlab = "Population (logarithmic)", # x-axis label
title = "Distribution of population worldwide", # title for the plot
subtitle = "Year: 2007", # subtitle for the plot
caption = "Data courtesy of: Gapminder Foundation", # caption for the plot
binwidth.adjust = TRUE, # adjust binwidth
binwidth = 1 # new binwidth
)
This shows the utility of gghistostats in case of data exploration.
Statistical analysis with gghistostats
Let’s say we are now interested in investigating whether the mean life expectancy in 2007 across the world has improved during the 20th-Century. In 1950, it was 48, so this is the test.value we are going to use.
Although there are still some countries where the life expectancy is low, on average, the life expectancy worldwide has improved compared to what it was in 1950.
gghistostats also provides the opportunity to compute Bayes Factors to quantify evidence in favor of the alternative (BF10) or the null hypothesis (BF01). In practice, you need to compute only one and the other will just be the inverse. In the current example, let’s say we want to quantify evidence in favor of the alternative hypothesis (H1) that the life expectancy in 2007 has improved significantly worldwide since 1957. The null, in this case, will of course be that there is no improvement.
gapminder::gapminder %>%
dplyr::filter(.data = ., year == 2007) %>%
ggstatsplot::gghistostats(
data = ., # data from which variable is to be taken
x = lifeExp, # numeric variable
messages = FALSE, # turn off messages
type = "bf", # bayesian one sample t-test
test.value = 48, # test value
xlab = "Life expectancy", # x-axis label
title = "Life expectancy worldwide", # title for the plot
subtitle = "Year: 2007", # subtitle for the plot
caption = "Note: black line - 1950; blue line - 2007", # caption for the plot
test.value.line = TRUE, # show a vertical line at `test.value`
centrality.para = "mean" # plotting centrality parameter (mean)
)
As this analysis shows, Bayes Factor value provides conclusive evidence in favor of the alternative hypothesis: Life expectancy worldwide has improved significantly since 1957.
Grouped analysis with gghistostats
What if we want to do the same analysis separately for all five continents? In that case, we will have to either write a for loop or use purrr, none of which seem like an exciting prospect.
ggstatsplot provides a special helper function for such instances: grouped_gghistostats. This is merely a wrapper function around ggstatsplot::combine_plots. It applies gghistostats across all levels of a specified grouping variable and then combines list of individual plots into a single plot. Note that the grouping variable can be anything: conditions in a given study, groups in a study sample, different studies, etc.
Let’s see how we can use this function to apply gghistostats for all five continents. We will be running parametric tests (one sample t-test, i.e.). If you set type = "np", results from non-parametric test will be displayed.
gapminder::gapminder %>%
dplyr::filter(.data = ., year == 2007) %>%
ggstatsplot::grouped_gghistostats(
# arguments relevant for ggstatsplot::gghistostats
data = .,
x = lifeExp,
xlab = "Life expectancy",
type = "p", # parametric test
test.value = 48, # test value against which sample mean is to be compared
test.value.line = TRUE, # show a vertical line at `test.value`
messages = FALSE, # turn off messages
centrality.para = "mean", # plotting centrality parameter (mean)
grouping.var = continent, # grouping variable with multiple levels
# arguments relevant for ggstatsplot::combine_plots
title.text = "Life expectancy change in different continents since 1950",
caption.text = "Note: black line - 1950; blue line - 2007",
nrow = 3,
ncol = 2,
labels = c("(a)","(b)","(c)","(d)","(e)")
)
As can be seen from these plots, life expectancy has improved in all continents in 2007 as compared to the global average of 1950. Additionally, we see the benefits of plotting this data separately for each continent. If we look at the standardized effect sizes (Cohen’s d), it is apparent that the biggest improvements in life expectancy outcomes were seen on the continents of Europe, Americas, and Oceania (just one data point is available here), while Asia and Africa exhibit the lowest improvements.
Although this is a quick and dirty way to explore large amount of data with minimal effort, it does come with an important limitation: reduced flexibility. For example, if we wanted to add, let’s say, a separate test.value argument for each continent, this is not possible with grouped_gghistostats. For cases like these, it would be better to use a function like purrr::pmap.