Experimental Design · Lesson 2 of 2
Evaluation
Evaluate whether results support the proposed relationship and whether the method could be biased. Suggest specific improvements: larger intervals, insulation, better alignment, zero correction or repeated readings, as appropriate.
By the end, you can
- Identify limitations.
- Distinguish random and systematic errors.
- Suggest realistic improvements.
Sources of error
An evaluation identifies what limited the experiment and how it could be improved, and the marks are for being specific. "Human error" and "be more careful" earn nothing, because they name no particular problem and no particular fix.
Start by classifying. Random errors scatter readings either side of the true value: reaction time on a stopwatch, judging when a ball passes a mark, small variations between repeats. They are reduced by repeating and averaging, by timing multiple oscillations, or by using electronic timing such as light gates. Systematic errors shift every reading the same way: a balance not zeroed, a ruler read at an angle, a stopwatch that runs slow. Averaging cannot remove them; they must be found and corrected, by checking the zero, by reading at eye level, or by calibrating against a known standard.
A zero error is the common systematic case: an instrument reading a value when it should read zero. Check it before starting and subtract it from every reading afterwards.
Then judge the evidence itself. Is the range wide enough to support the conclusion? Are there enough points? Do the repeats agree closely, which suggests good precision? Are there anomalies, and can they be explained? A conclusion should say what the data shows and how far it can be trusted, which is the honest version of the answer and the one that scores.
Worth remembering
- Name the specific error and the specific improvement; "be more careful" is worth nothing.
- Random error: scatter, reduced by repeats, longer intervals or electronic timing.
- Systematic error: a consistent shift, fixed by zeroing, calibration or reading at eye level.
- Judge the range, the number of points and the agreement between repeats.
Worked example
For a heat-capacity experiment, insulation reduces heat loss; merely repeating it does not remove the same systematic heat loss.
Investigate scatter and systematic error
Change a value and watch the readings follow. Open it on the lab stage
Investigate scatter and systematic error
There is no zero offset, and the readings lie up to 0.484 cm from the best-fit line.
- Furthest point from the best-fit line
- 0.484 cm
- 3.13 cm − 2.646 cm at 1 s
- Best-fit intercept
- 1.165 cm
- 1 cm true + 0 cm offset + 0.165 cm random
- Best-fit gradient
- 1.965 cm/s
- (20.815 cm − 1.165 cm) ÷ 10 s
Goal
Find out how random scatter and a zero offset each change a set of readings, and which of them taking more repeats can fix.
Predict, where you are
If Ithe, thewill
Say what you expect before you move anything. Check makes the change for you.
Test, still to come
Change one thing at a time and watch the readings. The notebook keeps a reading each time you record one.
Explain, still to come
After you check a prediction, this says why it happened.
Try this · 0 of 3
- Turn the random scatter up to 1.5 cm with no zero offset. Does the best-fit line still follow the true line?, not yet
- Set a zero offset of −1 cm with no scatter. How far does every reading move?, not yet
- With 1 cm of scatter and a 1 cm zero offset, take 10 repeats at each time. Which error does averaging reduce, and which does it leave?, not yet
reading = true value + zero offset + random error
at 5 s: 11 cm + 0 cm − 0.229 cm = 10.771 cm
Investigation notebook · 0 readings
Change one variable at a time and record the results. These readings stay here until you leave this lesson or reset the notebook.
About this model
Made-up readings of a distance that truly grows as 2 cm/s × time + 1 cm. Each reading’s random error comes from a fixed set drawn once, up to the scatter either side, so the same settings always give the same readings. The best-fit line is the least-squares straight line through the plotted points.
Practice
Warm-up · not marked
Check your understanding
Answer in your own words, then compare with the worked answer. Nothing here is saved or scored.
Worked answer
Time many oscillations and divide by their number; the reaction-time error becomes a smaller fraction of the total interval.
Check your method, units and explanation against it.
Marked practice
Set 1 of 4 · easiest first · Checked as you go, with the worked answer. Sign in or create a free account to save your progress.