Scatter diagrams
In a minute
A plots two measurements for each object as one cross, so you can see if they are linked. The pattern shows the : positive, negative or zero. A sums up the pattern and lets you estimate values.
You'll be able to
- Plot pairs of data on a scatter diagram.
- Read a scatter diagram and spot a point that doesn't fit.
- Name and describe positive, negative and zero correlation.
- Draw a line of best fit by eye.
- Estimate values from the line, and say when they are reliable.
Before you start2 quick questions on what this builds on
Revise first: 3.1 Coordinates
Drawing a scatter diagram
Sometimes you measure two things about each object, such as the distance a van drives and the fuel it uses. Is a longer drive linked to more fuel? A picture of all the pairs shows you.
1. Check the label on each axis. The first value of each pair usually goes across, but the labels decide.
2. Work out what one small square is worth on each axis. The scales are often different, and may not start at .
3. Plot each pair as a small cross (). Don't join the crosses.
Try it first, then reveal one step at a time.
- Across: one small square is km.
- km is km past the line, which is squares.
- Up: one small square is litres.
- litres is litre above the line, which is squares.
- Put a cross squares right of and squares above .
A cross far from the pattern is called an outlier. Here, one van used much less fuel than vans that drove about as far.
Try it first, then reveal one step at a time.
- Distance per litre is distance fuel. A van that used little fuel for its distance sits low, below the pattern.
- That is the cross at : km per litre.
- Check the next best: and are smaller.
- Round only at the end: .
Correlation
Look at the overall shape of the crosses. Do they drift up, drift down, or go nowhere?
1. Positive correlation: as one variable increases, the other tends to increase. The crosses rise from bottom left to top right.
2. Negative correlation: as one variable increases, the other tends to decrease. The crosses fall from top left to bottom right.
3. Zero correlation (or no correlation): there is no upward or downward trend.
"Tends to" matters: not every cross must follow the pattern. Crosses close to a straight line show strong correlation; loosely spread ones show weak correlation.
To describe it in context, say what happens to one variable as the other increases: "the further a van drives, the more fuel it uses".
Try it first, then reveal one step at a time.
- Plot them with width across. Most crosses rise from bottom left to top right: wider trees tend to be taller. This is positive correlation.
- Look for a cross far from the pattern. Trees and cm wide are m and m tall.
- The tree cm wide is only m tall, far below them. That is the storm-damaged tree.
If the pattern is hard to see, compare the mean -value of the left half of the points with the right half. Clearly higher on the right suggests positive, clearly lower suggests negative, and about the same suggests zero. It is only a rough check: look at the crosses too.
Try it first, then reveal one step at a time.
- In order of span, the scores are , then .
- Four smallest spans: mean score .
- Four largest spans: mean score .
- The means are equal, and the scores zigzag with no pattern: zero correlation.
Line of best fit
When there is correlation, one straight line can sum up the trend. You can read estimates from it.
1. Check the direction of the trend: up for positive correlation, down for negative.
2. Lay your ruler along the middle of the crosses. Move it until about as many crosses are above it as below, all along its length.
3. Draw one ruled line across the full spread of the data. It needn't pass through any cross, or through .
Ignore an outlier such as the storm-damaged tree. With zero correlation, no line fits.
Try it first, then reveal one step at a time.
- Go up from on the age axis to the line.
- Go across to the price axis and read .
- Check: the line falls dollars over months, dollars a month. So at months it is at .
Try it first, then reveal one step at a time.
- This time start on the price axis: go across from to the line, then down. Read months.
- Check: dollars at a month is months after , so .
- The data only goes from to months, so nobody knows if the trend carries on to .
- In fact the line gives dollars at months: impossible. Only trust estimates inside the range of the data.
Common mistakes
- Swapping the axes. Read each axis label before you plot.
- Misreading the scale, such as putting at . Work out what one small square is worth first.
- Joining the crosses dot to dot, or drawing a curve. A line of best fit is one ruled straight line.
- Getting the context backwards. Over the same route, a faster cyclist takes less time, so speed and time taken have negative correlation.
Exam tips
- "What type of correlation": one word is enough: positive, negative or zero.
- "Draw a line of best fit": one ruled straight line across all the crosses, balanced on each side. A curve or dot-to-dot line earns nothing.
- Estimating from your line: draw the up-and-across lines on the diagram. A correct reading from your own sensible line earns the mark.
Key takeaways
Practise this
22 past-paper questions on 9.5 in the 0580 Extended bank, each with a mark scheme and a worked explanation.