3 Flashcards
Stuff for the different ways of displaying data
Pie chart
These are most often used to show the relative sizes of the parts of whole. Pie charts are good for information visualisation because they can be considered aesthetically pleasing, however they are not considered to be scientifically accurate. Bar-charts are almost always recommended over pie charts because they are able to show data dispersion (e.g. standard deviation) as well as the measure of centrality (e.g. mean).
Bar-chart
A bar-chart is the most common way of representing comparative data when one of the comparators is categorical (nominal or ordinal) and the other is numerical. They are considered accurate and easy to interpret and are therefore highly recommended. They can be either horizontal or vertical but will always have two axes. Quite often it is best to just present these as 2D rather than 3D, which can be confusing.
Histogram
Visually a histogram is very similar to a bar-chart, with bars and two axes. However, there is an important difference: a histogram can have continuous numerical data on both of the axes but a bar-chart can only have continuous numerical data on one of the axes. The most common type of histogram is a frequency histogram, where the x-axis is continuous and the y-axis is counted quantitative.
Dot-plot
This is a graphical display of data using dots. It is very similar to bar-chart but the bars have been replaced by dots. The advantage of a dot-plot over a bar-chart is that it provides a better visual representation of the data dispersion but it is more suitable for smaller data-sets.
Box & whiskers
A conventional box & whiskers plot visually shows a 5-number summary of a data-set. Additional box & whiskers can be plotted alongside each other for comparative analysis. The 5-numbers summarised on a box & whiskers plot are:
lower extreme
lower quartile
median
upper quartile
upper extreme
Hence, the ‘box’ component of the plot represents the interquartile range and the whiskers represent the extremities. However, there are now several variations on the original format described, where the box & whiskers can represent different elements.
Box & whiskers diagrams can be used to summarise a single data-set (unlike the other graphs described thus far) and they are primarily used for non-parametric numerical data-sets rather than parametric numerical data-sets.
Scatter-plot
A scatter-plot is used to identify similarities between two data-sets rather than differences, or conventionally the relationship between two continuous numerical variables. A line can be added to the points to show a correlation and this can be a type of statistical test.
Other types of graph
Line graph: a diagram showing a line that joins several points and more often used to show the impact of time on a dependent variable. An advanced form of a line graph is a Kaplan-Mieir graph that is used for survival analysis.
Cumulative frequency curve: This is similar to a frequency histogram but uses a curve rather than bars and accumulates the numbers as we proceed along the x-axis. A variation on the cumulative frequency curve is a dose-response curve used in pharmacology, which you will have encountered.
Bubble plots: This is similar to a scatter plot, but the size of the bubble can represent a third variable. They are good for information visualisation but not always considered scientifically accurate.
Stem & leaf plot: these are similar to a frequency histogram or a dot-plot and display the general distribution of the data. They can be considered as a ‘hybrid’, somewhere between a table and a graph. However, they are only useful for moderately sized data-sets.
