Lesson 5 of 8 · 15 min
Describing a distribution
Describe what a display shows (the number of modes, the shape, the centre, the spread and any outliers) and interpret each feature in the context of the data.
What you will be able to do
- Identify the shape of a distribution as symmetric, positively skewed or negatively skewed, and count its modes.
- Estimate the centre and spread of a distribution from a display.
- Identify possible outliers and suggest reasons for them.
- Write a description of a distribution in the context of the data.
A display is only useful if you can say what it shows. Statisticians describe a distribution by the same four features every time, and so will you.
Key idea
Four features of a distribution
- Shape, including the number of modes (peaks).
- Centre: where the middle of the data sits.
- Spread: how far the data stretches out.
- Outliers: any values that stand apart from the rest.
Describe every feature in the context of the data, with numbers and units.
The three shapes. Skew is named after the side the tail stretches towards.
- Symmetric: the left and right sides are roughly mirror images, with the peak in the middle. Heights of adults of a similar age often look like this.
- Positively skewed: most values are bunched at the low end, with a long tail stretching towards the larger values (to the right). Travel times, incomes and house prices are typical.
- Negatively skewed: most values are bunched at the high end, with a tail stretching towards the smaller values (to the left). Scores on an easy test look like this.
Tip
Follow the tail
Skew is named after the tail, not the peak. A positively skewed distribution has its peak on the left, because the tail stretches out to the right, towards the larger (more positive) values.
In a positively skewed distribution, where is the long tail?
A mode of a display is a peak. Most distributions have one mode. A distribution with two clear peaks is bimodal, and it is usually a sign that two different groups have been mixed together, such as the heights of a group of young children and their parents.
These 30 sleep times are symmetric about 7.5 hours. Drag several values to the right to stretch out a tail, and watch the shape become positively skewed. Then try making it negatively skewed.
A fun run has a 5 km race and a 10 km race. The organisers draw one histogram of the finishing times of everyone in both races. What shape is the histogram most likely to have?
Centre is where the middle of the data sits. From a display, estimate the median (the value with half the data on each side) and give it with units. Spread is how far the data stretches: give the smallest and largest values, and say where most of the values lie. The next lessons calculate these exactly.
An outlier is a value that stands apart from the main body of the data, usually with a gap between it and the rest. When you find one, suggest a reason. It may be a recording error (a heart rate of 720 bpm was probably 72) or a genuinely unusual case (a student who lives 80 km out of town). Don't delete an outlier just because it is inconvenient.
Note
Coming in the next topic
Univariate data analysis 2 gives a rule, based on the quartiles, for deciding whether a value is an outlier. For now, judge by eye: a value well separated from the rest of the data is a possible outlier.
Worked example
Describe the distribution of parkrun finishing times
The histogram shows the finishing times of 80 people at a Brisbane parkrun, a 5 km run. Describe the distribution of the finishing times.
Step 1: Shape and modes
There is one clear peak, the class 25–<30 minutes, so the distribution has one mode. The columns fall away slowly to the right, with a long tail up to 55 minutes, but drop away quickly on the left. The distribution is positively skewed.
Skew is named after the tail, and this tail stretches towards the larger values.
Step 2: Centre
The median is the th time. Adding the columns from the left: times are under 25 minutes and are under 30 minutes. So the 40th time is in 25–<30 and the 41st is in 30–<35: the median is about 30 minutes.
With grouped data you can't find the exact median, but you can locate it.
Step 3: Spread
The times range from somewhere in 15–<20 minutes to somewhere in 50–<55 minutes, a spread of about 35 to 40 minutes. Most people ( of the 80) finished in 20 to 35 minutes.
Step 4: Outliers
There are no gaps between the columns and the tail thins out gradually, so no time stands apart. There are no clear outliers.
Step 5: Write it in context
Combine the four features into sentences about finishing times, with units.
Try it yourself first, then check the working one step at a time.
Answer
The finishing times are positively skewed with one mode (modal class 25–<30 minutes). The median is about 30 minutes, and the times spread from under 20 minutes to over 50 minutes, with most people finishing in 20 to 35 minutes. There are no clear outliers. The long tail of slower times is probably people who walked some or all of the course.
Method
Shape: count the peaks, then find the tail (to the right is positive skew, to the left is negative).
Centre: locate the median, or estimate the middle of the data.
Spread: give the smallest and largest values and where most of the data lies.
Outliers: look for values cut off from the rest by a gap.
Write each feature in a sentence with the variable, its units and the context.
Watch out for
Calling the distribution negatively skewed because the peak is on the left.
Describing only the shape, with no centre, spread, outliers or context.
The waiting times, in minutes, of 12 commuters at a ferry terminal were 4, 6, 7, 7, 8, 9, 9, 10, 11, 12, 13 and 41. Which time is a possible outlier?
Common mistake
Descriptions that lose marks
- Naming skew after the peak. Peak on the left, tail to the right: that is positive skew.
- No context. "It is skewed" earns little. "The travel times are positively skewed" says what is skewed.
- Calling the end of a long tail an outlier. A tail that thins out gradually has no outlier. An outlier stands apart, with a gap.
- No numbers or units. Write "centred at about 30 minutes", not "centred in the middle".
Exam tip
A model description
"The distribution of [variable] is [symmetric / positively skewed / negatively skewed] with [one mode / two modes]. The [variable] are centred at about [value and units] and range from [smallest] to [largest], with most between [a] and [b]. [There are no clear outliers. / The value [x] is a possible outlier, which could be because …]"
Summary
Describe a distribution by its shape (and number of modes), centre, spread and outliers, always in context.
Symmetric: mirror-image sides. Positive skew: tail to the right, towards larger values. Negative skew: tail to the left.
Two clear peaks (bimodal) usually mean two groups are mixed together.
An outlier stands apart from the rest of the data. Look for a reason before deciding what to do with it.
Quote numbers with units: "centred at about 30 minutes", not "in the middle".
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.