Ogives, variance and standard deviation, the shape of a distribution, and spotting outliers.
Practise Statistics in the app →
What gets asked
- Draw or read an ogive and take the median and quartiles from it.
- Calculate the variance and standard deviation of ungrouped data.
- Say whether data is symmetric or skewed, and in which direction.
- Identify outliers and comment on them in context.
You must be able to
- An ogive climbs through cumulative counts: read positions on the count axis first.
- Standard deviation is the square root of the mean of the squared deviations.
- Mean above median suggests a right skew; the box plot tells the same story.
- Fences at 1,5 IQR beyond the quartiles mark off the outliers.
Traps that cost marks
- Reading the median from the data axis instead of the halfway count. Find n over 2 on the cumulative axis, then trace across and down.
- Averaging the deviations without squaring them. Plain deviations cancel to zero — square first, average, then root.
- Calling the maximum an outlier automatically. Test it against the upper fence; a large value can still be ordinary.
Worked example
- Mean . Deviations: .
- Squares: ; mean of squares .
- .
Ogives and summaries
A running total, drawn as a curve, shows how many lie below any value.
- cumulative frequency
- A running total of the frequencies: how many values lie below the end of each interval.
- ogive
- A graph of cumulative frequency against the upper end of each interval: a cumulative frequency curve.
- median
- The middle of the ordered data: half the values lie below it.
- quartiles
- The lower quartile has a quarter of the values below it. The upper quartile has three quarters.
- histogram
- Touching bars, one per interval, each as tall as its frequency.
- box-and-whisker plot
- A box from to with a line at the median, and whiskers to the smallest and largest values.
Grouped data gives a frequency, or count, for each interval, such as . For cumulative frequencies, add each new frequency to the total before it. The total never restarts, and the last one is , the number of values.
To draw an ogive, plot each total above the upper end of its interval, starting at on the first lower end. Join the points with a smooth curve that only climbs.
For the median, start at position on the cumulative frequency axis. Go across to the curve, then down. Quartiles work the same way. To count the values below a value, go up, then across.
Choose a graph by its job. Counts below a value: an ogive. Each interval's count, and the shape: a histogram. Median and quartiles at a glance: a box-and-whisker plot. Shares of a whole: a pie chart, a circle cut into slices.
Rules to remember
- median at position ; at ; at (not on the information sheet)
Examples
Worked answer
- The first total is the first frequency: .
- Add each frequency to the total so far: , then .
- Then and . The last total is : a check.
Answer: ; ; ; ;
Each total counts every learner so far. It never restarts.
Worked answer
- With learners, the median is at position .
- Go across from to the curve, which passes between and .
- Then down to the time axis: the median is about minutes.
- is at position , and at .
- Across and down: is about minutes and about .
Answer: median about minutes; about ; about
All three start on the cumulative frequency axis and end on the time axis.
Worked answer
- Now start on the time axis at : go up to the curve, then across.
- is halfway from to , so the count is halfway from to : .
- About learners take less than minutes.
- At least minutes: learners.
Answer: about learners
The ogive counts learners below a time, so the rest are above it.
Traps
- Writing each interval's own frequency in the cumulative column. Each entry is a running total: , then , not .
- Starting on the horizontal axis instead of at the halfway count . Find position on the cumulative frequency axis, then trace across and down.
- Giving the position as the median. counts learners. The median is a time: about minutes.
- Choosing a histogram to show how many take under minutes. A histogram shows each interval's count. Counts below a value need an ogive.
Variance and standard deviation
Two learners can have the same average and still be nothing alike. One number shows who is steadier.
- mean
- The total of the values divided by how many there are, written .
- deviation
- How far a value is from the mean: . Below the mean, it is negative.
- variance
- The mean of the squared deviations, written .
- standard deviation
- The square root of the variance, written . It is in the units of the data.
The mean gives the centre of the data, not its spread. For spread, find each value's deviation from the mean. But the deviations always total : those below the mean cancel those above.
So square each deviation first, since a square is never negative. The mean of the squares is the variance, and its square root is the standard deviation. CAPS divides by , the number of values. Where a paper prints an information sheet, it gives and . Some print none, so learn both.
By hand, use a table of values, deviations and squares. For long lists, use the calculator's STAT mode. Read , which divides by . Never read , which divides by .
A small standard deviation means the values huddle close to the mean: the data are consistent. A large one means they are scattered, so bigger is not better.
Within one standard deviation of the mean means from to . Work out both ends, then count the values between them.
Rules to remember
- , where is the total of the values
- , so
Examples
Worked answer
- The times total , so minutes.
- Table of deviations and squares:
- The deviations total , as always. The squares total .
- Variance: . Standard deviation: minutes.
Answer: variance ; standard deviation minutes
Squaring stopped the cancelling, and the root brought the answer back to minutes.
Worked answer
- STAT mode for Sipho: and .
- For Lerato: and .
- Lerato's mean is higher, but Sipho's standard deviation is far smaller, so his marks are steadier.
Answer: Sipho's
Consistency is spread, so the standard deviations decide it, not the means.
Worked answer
- STAT mode: and .
- The ends: and .
- Between the ends lie ; ; ; ; ; ; : that is counts.
Answer: of the counts
The interval is centred on the mean, not the median, and built with , not .
Traps
- Averaging the deviations without squaring them. Plain deviations cancel to zero — square first, average, then root.
- Dividing by , or reading off the calculator. CAPS divides by : Thandi's goes over , not . Read .
- Calling Lerato more consistent because her mean is higher. A higher mean is a better average, not steadier marks. Compare the standard deviations.
- Building the interval with the variance instead of the standard deviation. Take the square root first: the ends are and .
Learn and practise “Variance and standard deviation” in the app →
Shape and outliers
Lopsided data and far-off values: how to spot each one, and why they matter.
- symmetric
- Balanced: spread out the same way on both sides of the median.
- skewed
- Lopsided: bunched up on one side, with a long tail on the other.
- interquartile range
- , the width of the box: the middle half of the data.
- fence
- A cut-off for outliers: below , or that far above .
- outlier
- A value beyond a fence, far from the rest of the data.
In symmetric data the mean and the median are equal, or very close. On a box-and-whisker plot, the median sits in the middle of the box and the whiskers match.
Skewed data have a tail, and the tail pulls the mean towards it. The median stays with the bunch, so it shows a typical value better. Name the skew after the tail. A tail of high values means skewed to the right, or positively skewed. A tail of low values means skewed to the left, or negatively skewed.
On a box-and-whisker plot, compare the two parts of the box, and . Compare the two whiskers too. The longer side is the tail's side.
For outliers, build a fence interquartile ranges beyond each quartile: below and above . Only a value beyond a fence is an outlier. No information sheet gives the fences, so learn them. In context, an outlier may be a recording slip, or a real but unusual case.
Rules to remember
- lower fence ; upper fence
Examples
Worked answer
- Left part of the box: . Right part: .
- Left whisker: . Right whisker: .
- Both are longer on the left, so the tail is there: skewed to the left.
Answer: skewed to the left (negatively skewed)
The few low marks form the tail, and the tail names the skew.
Worked answer
- The median is the 6th time: minutes.
- The times total , so the mean is minutes.
- The mean is above the median, pulled up by the and the : skewed to the right.
Answer: skewed to the right (positively skewed)
The tail pulls the mean, not the median.
Worked answer
- , the median of the lower five times. , the median of the upper five.
- IQR: , and .
- Lower fence: . No time lies below it, so is not an outlier.
- Upper fence: . Only lies above it, so is an outlier.
Answer: minutes is an outlier; and minutes are not
Each end met its own fence. In context, that learner may live far away.
Traps
- Naming the skew after the side where the values bunch up. Name it after the tail. Marks bunched high with a low tail are skewed to the left.
- Calling the data symmetric because the mean and the median are both averages. When they differ, the data are skewed, as the travel times are.
- Building the fences from the median instead of the quartiles. Fences hang off the quartiles. From the median, would wrongly flag .
- Calling the maximum an outlier automatically. Test it against the upper fence; a large value can still be ordinary.