1.                                                                                                                                     [10 marks]

(a)

[5 marks]

NOTE that this would require a geom_col rather than a geom_bar because the data set already contains the counts. Also note that the x-axis scale is x_continuous, even though it represents a date, because the values are just week numbers.

geom:

col

aesthetics:

x  =  Week

y  =  Count

stat:

identity

coord:

cartesian

scale:

x  =  x_continuous

y  =  y_continuous

(b)

There is no need to provide code to generate the data set.

library(ggplot2)

ggplot(scotland)  +

geom_col(aes(x  =  Week,  y  =  Count))  +

ggtitle("Deaths per  week  involving  Covid-19")  + ylab(NULL)  +

scale_x_continuous(breaks  =  12:37)  +                      scale_y_continuous(expand  =  expansion(add  =  0))

[5 marks]

2.                                                                                                                                     [10 marks]

The legend can be drawn using a layout of 3 rows and 2 columns, with the left column being fixed width (e.g., 1 inch) and the right column filling the remaining space ("null" units).

We push a viewport with that layout and then a subsequent viewport for each “cell” to draw the individual components of the legend.

In the first column, there are two rectangle grobs, one filled with a dark blue and the other grey, plus a segments grob, with a thick dashed line style.

The text is all left-justified.

This is a fair bit of code, but it has a simple, repetitive structure.

library(grid)

grid.newpage()

blue  <-  "#284f99"

grey  <-  "#b9b9b9"

lay  <-  grid.layout(3,  2,

widths  =  unit(c(1.25,  1),  c("in",  "null")), heights  =  unit(2,  "lines"))

pushViewport(viewport(layout  =  lay))                                                   pushViewport(viewport(layout.pos.col  =  1,  layout.pos.row  =  1)) grid.rect(width=.8,  height=unit(1,  "lines"),

gp=gpar(col=blue,  fill=blue))

popViewport()

pushViewport(viewport(layout.pos.col  =  1,  layout.pos.row  =  2)) grid.rect(width=.8,  height=unit(1,  "lines"),

gp=gpar(col=grey,  fill=grey))

popViewport()

pushViewport(viewport(layout.pos.col  =  1,  layout.pos.row  =  3)) grid.segments(.1,  .5,  .9,  .5,

gp=gpar(lwd=3,  lty="dashed",  lineend="butt"))

popViewport()

pushViewport(viewport(layout.pos.col  =  2,  layout.pos.row  =  1))            grid.text("Deaths  involving  COVID-19",  x=unit(5,  "mm"),  just="left") popViewport()

pushViewport(viewport(layout.pos.col  =  2,  layout.pos.row  =  2))    grid.text("Other  causes  of  death",  x=unit(5,  "mm"),  just="left") popViewport()

pushViewport(viewport(layout.pos.col  =  2,  layout.pos.row  =  3))

grid.text("Average  deaths per  corresponding  week  over previous  5  years", x=unit(5,  "mm"),  just="left")

popViewport()

popViewport()

It is also acceptable to position each element explicitly as long as sensible units are used.

3.                                                                                                                                     [10 marks]

(a)

[3 marks]

Redundant coding is used in both the barplot and the multi-panel plot by adding text labels.  Each bar in the barplot represents a number of deaths by its height and by the text label on the bar. Each panel in the multi-panel plot represents the relative deaths due to  COVID-19 versus Other causes using shaded areas and a text label describing the percentages.

(b)                                                                                                                                 [4 marks] If we overlapped the areas representing COVID-19 and Other deaths, we would create a problem of not being able to see both areas at once. We would at least have to draw the COVID-19 deaths on top of the Other deaths.  We could also just draw a line (rather than filling an area) for each source of deaths and/or we could use a semitransparent fill for each area.  Finally, we could use (more) multi-panel figures, producing a panel for COVID-19 deaths and another panel for Other deaths, for each age group. I think that the last option would be the most effective.

(c)                                                                                                                                 [3 marks] Yes (as far as we can tell in the barplot), the Principle of proportional ink is being obeyed in both plots because the y-axis scales start at zero. This means that the height of the bars in the bar plot and the height/area of the regions in the multi-panel plot correspond to the data values that they represent (number of deaths in a day in the barplot and number of deaths/total deaths from each source in the multi-panel plot).  The width of the bars in the bar plot is redundant (so the data-ink ratio is not the best), but because it is consistent that does not cause a major problem.

4.                                                                                                                                     [10 marks]

(a)

Visual tasks:

[5 marks]