Pseudoreplication: When n Is Smaller Than It Looks · seed 1 · A4, ink-friendly. The answer key prints on its own page for grown-ups.

Counting only truly independent replicates

Science · Scientific Inquiry · ages 22-23
Name ______________________   Date ____________

  1. What is the honest first step with technical repeats?

    • Delete them to save space
    • Average them to one value per unit
    • Multiply them to raise the sample size

  2. What does pseudoreplication mean?

    • Treating dependent repeats as independent evidence
    • Repeating an experiment in a second laboratory
    • Preregistering the analysis plan in advance

  3. Twelve cells from three treated wells give an honest n of twelve.

    Circle one:   True   False

  4. Readings nest within patients and litters within mothers. What analysis does this demand?

    • One that ignores all grouping and pools everything
    • One that reports only the largest cluster
    • One that models variation at the unit level of the hierarchy

  5. Why does pretending clustered points are independent shrink error bars?

    • The math divides by an inflated count while real uncertainty stays put
    • Clustering always removes all measurement noise
    • Error bars grow whenever n is overstated

  6. A legend reports n equals 12 with points from 3 named animals. What do you suspect?

    • Exemplary replication at every level
    • Nesting: readings counted as units, so precision is borrowed
    • A typo that changes nothing

  7. Students in one classroom are counted as independent trials of a teaching method. What is wrong?

    • Nothing; classrooms never share influences
    • Teaching methods cannot be studied at all
    • Classmates share hidden causes, so the effective sample shrinks toward one cluster

  8. A colleague rebuilds the analysis at the unit level and significance vanishes. What happened?

    • The rebuild destroyed perfectly good data
    • The original result was borrowed precision from clustered points
    • Unit level analyses always erase true effects
LightMySky · lightmysky.comW1-mt_2gRWB6RTaI-s1

Answer key

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

Counting only truly independent replicates W1-mt_2gRWB6RTaI-s1

  1. Average them to one value per unit · One mean per unit keeps the error math honest.
  2. Treating dependent repeats as independent evidence · Echoes of one unit add no fresh information about the treatment.
  3. False · The treatment touched three wells, so the honest n is three.
  4. One that models variation at the unit level of the hierarchy · Nesting demands a hierarchy with units above and readings below.
  5. The math divides by an inflated count while real uncertainty stays put · Sampling error shrinks with the independent count, so honest counting decides honest bars.
  6. Nesting: readings counted as units, so precision is borrowed · Legends betray the problem when points nest inside far fewer units.
  7. Classmates share hidden causes, so the effective sample shrinks toward one cluster · When units cluster, each new one brings an echo rather than fresh information.
  8. The original result was borrowed precision from clustered points · Honest n restores honest uncertainty, and noise stops looking like signal.
Worksheet · LightMySky