Graphs and data analysis
In a minute
A graph turns a table of results into a relationship you can see, test and measure. This lesson covers , and , plotting with a , , , and .
You'll be able to
- Tell direct and inverse proportionality from positive and negative correlation.
- Sketch a graph with labelled but unscaled axes to show a trend.
- Plot graphs with sensible scales and draw a line or curve of best fit.
- Find rates of change from gradients and tangents, and interpolate and extrapolate.
- Draw uncertainty bars and use them to find the in a gradient.
- Use logarithmic scales, and linearise a graph when theory predicts its form.
Before you start3 quick questions on what this builds on
Revise first: 1.1 Describing motion, 1.2 Equations of motion and motion graphs and 6.3 Uncertainties and errors
Relationships and sketches
Before you plot anything, ask how should change when changes.
A straight line missing the origin, with , is linear but not proportional.
Proportionality is an exact rule; a correlation is any trend. Even a strong correlation doesn't show cause, because a third factor may drive both variables.
A **sketch** shows the shape, not the values. Label both axes with the quantities but give no scales. Show where the line starts and whether it is straight, steepening or flattening.
Try it first, then reveal one step at a time.
- If , the ratio is the same for every sample.
- , and .
- The ratio is constant, so . The constant is the density, in .
Try it first, then reveal one step at a time.
- Find the rule first: , so .
- So : zero at , and doubling multiplies it by 4.
- Label the axes (vertical) and , with no scales. The curve starts at the origin, because the ball starts from rest. It curves upwards, getting steeper, because is proportional to .
Plotting a graph and drawing the best fit
To **plot** a graph, mark each point accurately on a scaled grid:
1. Put the variable you change on the -axis and the one you measure on the -axis.
2. Label each axis as quantity / unit: / s, / N.
3. Choose scales so the points fill more than half the grid each way. Use steps of 1, 2 or 5 per square, times a power of ten.
4. Include the origin only if you need it, such as to read an intercept.
5. Mark each point with a small, neat cross.
Rule a straight line, or draw one smooth curve if the points clearly curve. Never join the dots: the line averages out the scatter.
Don't force the line through the origin. Let the data decide where it crosses the axis.
A point far from the trend is an anomalous result. Check or repeat it; if it still doesn't fit, leave it out of the line and say why.
Try it first, then reveal one step at a time.
- Divide the largest value by the squares: A per square.
- Round up to a step of 1, 2 or 5 times a power of ten: per square.
- Use a scale of 0.05 A per large square. The largest reading then reaches 9.2 squares, so the points fill most of the grid. A scale of 0.1 A per square would use less than half of it. Label the axis / A.
Try it first, then reveal one step at a time.
- Deal with the odd point first, then fit the rest.
- First check or repeat the odd reading, because it is probably an anomalous result. If it still lies far from the trend, leave it out of the line and label it anomalous. Use a ruler to draw one straight line with the other six points scattered evenly above and below it. Don't join the dots or force the line through the origin.
Gradients and reading values off a graph
A rate of change, how fast changes with , is the of the graph.
For a straight line, draw a large triangle using two points on the line that are far apart, not data points off it.
Its unit is the unit over the unit: on a velocity–time graph, .
On a curve, the rate changes from point to point. Draw the tangent, a straight line touching the curve at that point. Its gradient is the rate of change there.
The multiplies the axes instead: velocity times time gives displacement.
**Interpolation** is reading a value between your data points. **Extrapolation** is extending the line beyond them, for example back to an axis to find an intercept.
Extrapolation assumes the trend continues outside the measured range, and it may not: an overstretched spring changes behaviour.
Try it first, then reveal one step at a time.
- Acceleration is the gradient of the tangent.
- and .
- .
Try it first, then reveal one step at a time.
- Gradient: .
- Going back from to , the length falls by .
- . This is extrapolation: is outside the range used.
Uncertainty bars
A graph shows each point's with a bar.
A point at has a vertical from to . Draw bars on each axis with a noticeable uncertainty.
A good line of best fit passes through every bar. If no straight line can, the data are not linear, some points are anomalous, or the uncertainties were underestimated.
For the uncertainty in a gradient, draw the steepest and shallowest straight lines through every bar:
An intercept's uncertainty is found the same way, from the two lines' intercepts.
Data are consistent with direct proportionality only if some straight line through the origin passes through every bar.
Try it first, then reveal one step at a time.
- .
- Give it to 1 significant figure, and round to match: .
Try it first, then reveal one step at a time.
- Lines that fit the bars cross the -axis between and . Is zero in that range?
- Direct proportionality needs a straight line through the origin that passes through every uncertainty bar. The lines that fit the bars have intercepts from to , a range that excludes zero. So the data are not consistent with being directly proportional to .
Logarithmic scales
Some data span many powers of ten, such as star luminosities. On an ordinary axis the small values crowd into one corner.
Each factor-of-10 step is a decade. Zero never appears on a .
The position of a value along the axis is set by . So a point a fraction of the way from to has the value .
So halfway between 100 and 1000 is , not 550. Minor gridlines crowd towards the next power of ten.
To construct one, use log paper, or plot on an ordinary axis, labelled like .
If both axes are logarithmic and the points are straight, then , with gradient .
The HR diagram in 5.13 uses log scales.
Try it first, then reveal one step at a time.
- The gridlines are and Hz, so the peak is at Hz.
- Not : that treats the axis as linear.
Try it first, then reveal one step at a time.
- and .
- .
- So , which means : Kepler's third law, from 4.1.
Linearising a graph
Is a curve or ? You can't tell by eye, but a straight line is easy to test.
Match the equation to : what multiplies the -axis quantity is the gradient.
1. : plot against . The gradient is .
2. : plot against . The gradient is .
3. with unknown: plot against . The gradient is .
A mass on a spring oscillates with period (3.1). Squaring gives , so against is straight with gradient .
Linearise only when theory gives an equation to test. Data that are already straight need no change.
Try it first, then reveal one step at a time.
- From , the gradient is .
- , to 3 significant figures like the gradient.
- Check the unit: .
Try it first, then reveal one step at a time.
- Match to .
- Plot against , because the theory then has the form . If the theory is right, the points lie on a straight line through the origin, within their uncertainty bars. The gradient of that line is .
Common mistakes
- Forcing the line through the origin. Draw the best line for the points, then test the origin against the bars.
- Taking a gradient from data points off the line, or close together. Use two far-apart points on the line.
- Leaving the unit off a gradient. Divide the unit by the unit: over gives .
- Joining the points dot to dot. Draw one smooth line or curve through the trend.
- Reading a logarithmic axis as if it were linear. Halfway between 10 and 100 is , not 55.
Exam tips
- "Determine the gradient": show a large triangle and the two points you read, then the value with its unit.
- "Sketch": labelled axes and the right shape, no scale. "Plot" and "draw": accurate, on a scaled grid.
- "Deduce whether the data support" proportionality: say whether a line through the origin passes through every bar, then conclude.
- "Outline why" a graph is straight: rearrange into , name each axis, and say what the gradient equals.
Key takeaways
Practise this
21 past-paper questions on Graphs, data processing and modelling in the Physics HL bank, each with a mark scheme and a worked explanation.
Also: Experimental design, investigation and evaluation (16)