Method Validation: Accuracy, Precision, Recovery and Robustness · seed 1 · A4, ink-friendly. The answer key prints on its own page for grown-ups.

Prove the numbers, not just make them

Science · Chemistry · ages 21-22
Name ______________________   Date ____________

  1. Which experiment tests precision?

    • Buying a certified reference material
    • Running repeats and watching their spread
    • Publishing the method online

  2. Which experiment tests accuracy?

    • Measuring a sample of known content
    • Repeating the same sample ten times
    • Changing the room temperature

  3. Repeatability and reproducibility mean the same thing.

    Circle one:   True   False

  4. A method gives tight repeats but 78 per cent recovery on certified material. What is wrong?

    • Nothing, precision is all that matters
    • Random noise: the repeats are too close
    • A systematic loss: precise results with a biased low answer

  5. You spike 10 mg into a blank and recover 7.8 mg. What is the recovery in per cent? Type the number.

    Answer: ______________

  6. How do you check robustness of a method?

    • Vary one parameter at a time and read the effect
    • Change everything at once and hope
    • Run it a single time very carefully

  7. A lab reports repeatability only and calls the method validated. A reviewer objects. What is missing?

    • More decimal places on the repeatability
    • Reproducibility in other labs and robustness to small changes
    • A longer method name

  8. Recovery is 78 per cent and precise. Which follow-up experiment confirms a loss during extraction?

    • Repeat the same run five more times
    • Buy a newer instrument
    • Spike before versus after extraction and compare the two recoveries
LightMySky · lightmysky.comW1-mt_ihmF2fl4OC-s1

Answer key

For grown-ups. Fold this page away before handing over the rest.

Prove the numbers, not just make them W1-mt_ihmF2fl4OC-s1

  1. Running repeats and watching their spread · Precision is agreement among repeats, nothing else.
  2. Measuring a sample of known content · Only known content reveals closeness to the truth.
  3. False · One is same lab, the other is different labs, and both are reported.
  4. A systematic loss: precise results with a biased low answer · Agreement plus a low known-sample result means bias, not noise.
  5. 78 · 7.8 divided by 10 is 0.78, which is 78 per cent.
  6. Vary one parameter at a time and read the effect · One-at-a-time changes reveal which wobble matters.
  7. Reproducibility in other labs and robustness to small changes · Same-lab agreement alone never proves the method travels or endures wobble.
  8. Spike before versus after extraction and compare the two recoveries · If post-extraction spikes recover fully, the extraction step is where sample dies.
Worksheet · LightMySky