In gentrywhite/MXB107: Unit Package for MXB107

  # if problems with knitting directory:
  # set Knit Directory = Document Directory

  # install libraries needed for worksheet
knitr::opts_chunk$set(
  fig.align = 'center',
  collapse = TRUE,
  comment = "#>"
)

  library(MXB107)
  library(learnr)

htmltools::includeHTML("header.html")

Bivariate Data

Multivariate data arises when we collect more than one observations per experimental unit, for example surveys can ask individuals multiple questions, resulting in a multivariate data set. Bivariate data is a special case where we have two measurements per experimental unit (note that bivariate data can be extracted from a larger multivariate data set for the purposes of analyses). Bivariate data is useful especially when we want to examine the nature of the relationship between two variables. Bivariate data can be categorical, numerical, or some combination of the three. This week we will look at means of summarising categorical data to explore relationships between the two variables.

Bivariate Categorical Data

Bivariate cetegorical data occurs when we have two observations of an experimental unit that are both categorical variables. These can be summarised easily and compactly in tables, which can provide an easy to read and reference that numerically sumarises the data, but also provides some structure to indicate the relationship between the variables.

Tables

Tabular representations of bivariate data are constructed by assigning rows and columns to each of the categories for the two variables, the cells of the table then contain the counts (or sometimes proportions) of individuals who are in both categories. These kind of tables are often called contignecy tables and are common tools of data analysis. In R the command table()

The is also a function in the knitr package (which is used to create this document called) kable that can be used to create tables in a variety of formats, including HTML, which can look nicer in your Rmd documents. Use the episodes data and create a table showing how many shows of each series pass or fail the Bechdel-Wallace Test.

:::{.boxed} Hint: The kable package can be used to create tables. You will need to use group_by(), count() and pivot_wider() in addition to kable() to create a nice HTML table. :::

 

data("episodes")

  episodes%>%
    group_by(Series.Name,Bechdel.Wallace.Test)%>%
    count(Series.Name, Bechdel.Wallace.Test)%>%
    pivot_wider(names_from = Series.Name, values_from = n)%>%
    kable()

Note that the frequencies can be deceiving because series have different numbers of episodes. Instead it is more useful to create tables of proportions. Update the table above to show the proportions rather than counts.

:::{.boxed} Hint: The process for creating this table is similar to the table of counts, except in this case we will use mutate() to create a new variable called prop for the proportions. The digits argument in kable() controls the significant digits displayed. :::

 

data("episodes")

 episodes%>%
    count(Series.Name, Bechdel.Wallace.Test)%>%
    group_by(Series.Name)%>%
    mutate(prop = n/sum(n))%>%
    select(-n)%>%
    pivot_wider(names_from = Series.Name, values_from = prop)%>%
    kable(digits = 2)

Graphical Summaries

These proportions and counts can visualised in a single plot as well. Use the episodes data to create a bar plot to display the counts of episodes that pass and fail the Bechdel-Wallac Test for each series.

:::{.boxed} Hint: We will use ggplot() to create plots like we did in Workshop 1 but we will need to add the dodge option in geom_bar() to compare results for each series side-by-side. :::

 

data("episodes")

ggplot(episodes,aes(x = Series.Name,fill = Bechdel.Wallace.Test))+
  geom_bar(stat = "count",position = "dodge")+
  xlab("Series")+
  ylab("Number of episodes")+
  ggtitle("Star Trek episodes by Series that Pass or Fail the Bechdel-Wallace Test")+
  theme(plot.title = element_text(hjust = 0.5),legend.position = "bottom")

Notice that this plot is actually a little "busy" and it is difficult to see the differences across the series. Update the plot above as a "stacked" bar plot.

:::{.boxed} Hint: eliminating the dodge option creates a "stacked" plot. :::

:nbsp&

data("episodes")

ggplot(episodes,aes(x = Series.Name,fill = Bechdel.Wallace.Test))+
  geom_bar(stat = "count")+
  xlab("Series")+
  ylab("Number of episodes")+
  ggtitle("Star Trek episodes by Series that Pass or Fail the Bechdel-Wallace Test")+
  theme(plot.title = element_text(hjust = 0.5),legend.position = "bottom")

Bivariate Combined Data Types

Combined bivariate data includes both numeric and categorical data, these can be very common datasets to find, and arise from looking at how different factors influence some numerical measurements.

Summary Tables

As in the case of purely categorical data, tables can be useful for numerically summarising results. Use the episodes data to create a table showing the average IMDB ranking for each series.

:::{.boxed} Hint: Use summarise() to compute the mean IMDB ranking by series. :::

:nbsp;

data("episodes")

  episodes%>%
  group_by(Bechdel.Wallace.Test)%>%
  summarise(mean_imdb = mean(IMDB.Ranking))%>%
  kable()

But numerical summaries like the mean or median give limited information, so often time graphical summaries can tell us more.

Graphical Summaries

As we have seen previously, side-by-side graphical depictions like historgrams or boxplots can be very useful for illustrating differences between data sets. Use teh episodes data to create a side-by-side histograms comparing the IMDB rankings for episodes that pass or fail the Bechdel-Wallace Test.

:::{.boxed} Hint: Use facet_wrap() to create side-by-side plots. :::

library(MXB107)
data("episodes")