25.8 - Evaluating a Method - Difficulties, Errors & Improvements
What an evaluation question looks like
Evaluation questions ask you to identify what makes an experiment difficult to carry out accurately, or how to improve it.
The three forms
- One mark, difficulty only - e.g. "Suggest one practical reason why it is difficult to obtain accurate values for and for . [1]"
- Two marks, difficulty and improvement - Answered on separate lines: "Describe one difficulty... Suggest an improvement to the apparatus to overcome this. [2]"
- Two marks, phrased as error - "Describe one possible source of inaccuracy in the method... and suggest one improvement to reduce its effect. [2]"
How they're marked
- The difficulty has to be specific to the method shown.
- The improvement has to fix that particular difficulty.
- A general answer scores no marks.
- The question is never asking for a mistake the student made. The method is assumed to have been carried out properly - you're being asked what's hard about it even when it's done well.
A typical question
Describe one difficulty that is experienced when using a ruler to measure the height of the image in this experiment. Suggest an improvement to the apparatus to overcome this.
difficulty ..............................................
improvement ......................................... [2]
Finding the difficulty
A difficulty that scores is a specific reason why a particular measurement in this method is hard to get right.
Three questions that find one
- Is there something you have to judge by eye? E.g. deciding when a metre rule is exactly balanced, when an image is in focus, or when two pins line up.
- Is the thing you're measuring poorly defined? E.g. pencil lines have thickness, pins have a diameter, and light beams have fuzzy edges.
- Is something moving, changing, or in the way? E.g. water cools while reading a thermometer, a wire warms up during a current reading, or hands block the light.
Recurring difficulties by experiment
| Experiment | Difficulties that score |
|---|---|
| Balancing a metre rule | Hard to judge the exact balance point (the rule looks level over a small range). Hard to judge the centre of the load. The load or rule can slide. |
| Ray-tracing with pins | Hard to line up the pins by eye. The pins are thick compared with the ray. The drawn lines are thick. The protractor reads only to the nearest degree. |
| Lens and screen | Hard to tell exactly when the image is in focus. Your hand or the ruler blocks the light while you measure. The image edges aren't sharp. |
| Timing oscillations | Reaction time at start and stop. Judging when the swing reaches the end of its travel. The oscillations die away. |
| Thermometer and cooling | The temperature is still changing as you read it. The thermometer takes time to reach the liquid's temperature. |
| Resistance of a wire | The wire heats up as current passes, changing its resistance. Hard to fix the crocodile clip at exactly the right point on the scale. |
Suggesting an improvement
- The improvement mark is only given if the improvement addresses the difficulty you named.
- State the difficulty first, then give the fix for that specific problem.
Matching the fix to the problem
| Difficulty | Improvement that scores |
|---|---|
| Hard to judge when the rule balances | Use a knife-edge pivot rather than a broad one, so the balance point is sharper |
| Hard to judge the centre of the load | Hang the load from a thin loop or hook at a marked point, rather than resting it on the rule |
| Your hand or ruler blocks the light | Use a translucent screen and measure the image from behind, or fix a printed grid to the screen |
| The pins are thick and hard to align | Use thinner pins, place them further apart, and view the bottoms of the pins |
| The drawn lines are thick | Use a sharper pencil |
| Reaction time when timing | Time 20 oscillations and divide, rather than timing one |
| The wire heats up | Switch the current off between readings, or use a lower current |
| The temperature is still changing as you read | Stir, and wait until the reading is steady before recording it |
Random and systematic error
- Random error - Varies from reading to reading in both directions (e.g. reaction time, judging a level, estimating between scale divisions).
- Taking several readings and calculating an average reduces random error because the errors partly cancel.
- Systematic error - The same size and direction every time (e.g. a zero error on a meter, a pivot not at the centre of mass, a worn ruler end).
- Averaging does not reduce systematic error. The cause must be removed: zero the instrument, balance the bare rule first, or measure from the 1 cm mark.
- Identifying them in results - If values scatter on both sides of the expected value, the error is random; if they're all shifted in one direction, it's systematic.
Comparing which value is more reliable
- A fixed reading error is a larger percentage of a small measurement than of a large one.
- Therefore, a value calculated from larger measurements is more reliable (e.g. timing 20 oscillations instead of 1, or using a large gradient triangle).
Common exam questions
Question 1
A student answers an evaluation question with "human error". Why does this score nothing?
Because it does not say what was hard about this particular method. What is usually meant by it is something specific and creditworthy, such as judging by eye when the metre rule is balanced, and writing that specific thing is what earns the mark.
Question 2
Describe one difficulty in finding the balance point of a metre rule on a pivot.
The rule appears level over a small range of positions rather than at one exact point, so the balance point has to be judged by eye and different people would judge it differently.
Question 3
A student measures the height of an image on a screen with a ruler. Describe one difficulty and suggest an improvement.
The difficulty is that the ruler and the student's hand block the light from the lens, so the image is obscured while it is being measured. The improvement is to use a translucent screen and measure the image from behind it, or to fix a printed grid to the screen and read the height off that.
Question 4
What is the difference between random error and systematic error?
Random error varies in size and direction from reading to reading, so repeating and averaging reduces it. Systematic error is the same size and in the same direction every time, so averaging does not help and the cause has to be found and removed.
Question 5
Give one example of a systematic error and say how you would deal with it.
A force meter that reads 0.2 N with nothing hanging from it. Either adjust it to read zero before starting, or subtract 0.2 N from every reading afterwards.
Question 6
Why is "repeat the readings and take an average" a poor answer to most improvement questions?
Because it only helps with errors that fall differently each time. If the same thing goes wrong in the same direction on every reading, repeating produces the same wrong answer more precisely.
Question 7
In a set of results, how can you tell whether an error is random or systematic?
Look at which way the mismatches go. If they scatter either side of the expected value, the error is random; if they are all in the same direction, it is systematic.
Question 8
Two values of a focal length are calculated, one from small measurements and one from large ones. Which is more reliable, and why?
The one from the larger measurements. The reading error is roughly the same size in both cases, so it is a smaller percentage of a large measurement than of a small one.
Question 9
Suggest one difficulty in a ray-tracing experiment with optics pins.
It is hard to line the pins up by eye, and the pins have a thickness of their own, so the line drawn through them may not be exactly along the ray.