Experimental design worksheet with answer key

8 problems on experimental design, ready to print for class or homework. The answer key prints on its own page, so hand out everything before it.

Problem 1

A gardener tests whether a new LED grow light changes basil growth. She has 36 basil seedlings of the same variety and age. Twelve are randomly assigned to a standard fluorescent light, 12 to the LED at low intensity, and 12 to the LED at high intensity. All 36 get the same pot size, soil, watering schedule, and room temperature. After 4 weeks she records each plant's height in centimeters. (a) Name the experimental units, the explanatory variable, and the response variable. (b) How many treatments are there, and how many units receive each one? (c) Which of the four elements of a well-designed experiment do the shared soil and watering supply? (d) Can a difference in mean height be attributed to the lighting?

Problem 2

A high school compares GPAs across its 480 students. The 180 who play a school sport average a GPA of 3.35, and the 300 who do not average 3.11. The school newspaper reports that playing a sport raises GPA. (a) Name the explanatory and response variables. (b) Is this an experiment or an observational study, and how can you tell? (c) Verify the school's stated overall mean GPA of 3.20 and state the gap between the two groups. (d) Give a confounding variable and explain both links that make it confounding. (e) What would have to change before the newspaper's causal claim is justified?

Problem 3

A pharmaceutical company tests a new allergy tablet on 84 adult volunteers who have seasonal allergies. Describe a completely randomized design that compares the tablet with a placebo, using a symptom score from 0 to 24 recorded after 2 weeks as the response, where a lower score means fewer symptoms. Then say what the placebo and the blinding each accomplish.

Problem 4

A clinic tests a knee-pain cream. Sixty patients are randomly assigned, 30 to the active cream and 30 to an identical inert cream. Patients are not told which cream they received. The physical therapist who scores each patient's knee function at week 6 does know the assignments. (a) Is this study single-blind or double-blind? (b) Name one specific way the therapist's knowledge could distort the results. (c) State the fix. (d) The clinic director asks why the control group could not simply receive nothing. Answer him.

Problem 5

A coach compares two 8-week training programs using 60 runners. Thirty-six of them have finished a marathon before and 24 have not, and prior experience is known to have a large effect on finish time. The response is 10K finish time in minutes at the end of the 8 weeks. (a) Explain why a randomized block design is a better choice here than a completely randomized design. (b) Describe the design, including how the randomization runs and the size of every group. (c) How many runners follow each program in total? (d) Does blocking change what the coach can conclude about cause and effect?

Problem 6

Eight volunteers each type the same passage on two keyboard layouts, A and B. For each volunteer a coin flip decides which layout is used first, and a 10-minute rest separates the two attempts. Typing speeds in words per minute:

Volunteer12345678
Layout A6258714966546068
Layout B6557765468586371

(a) Name the design and say what the pairing controls. (b) Describe the random assignment and count how many equally likely order assignments there are. (c) Find the eight differences (B minus A) and their mean. (d) What does the coin flip protect against?

Problem 7

A greenhouse trial crosses 2 fertilizer brands with 3 watering schedules on 48 identical pots of the same seedling. The greenhouse has a bright south half and a shaded north half, 24 pots in each half, and light has a strong effect on growth. (a) How many treatments does crossing the two variables produce? (b) Set up a randomized block design that uses the greenhouse halves as blocks, and state how many pots receive each treatment within a block and overall. (c) Describe the randomization inside one block. (d) Is light a confounding variable in this design? Explain.

Problem 8

A researcher recruits 20 volunteers from a fitness app's mailing list and randomly assigns 10 to a new 6-week strength routine and 10 to the app's standard routine. Two of the 20 have lifted competitively before, and that experience is expected to affect the response, a one-rep-max bench press in kilograms. (a) How many different sets of 10 volunteers could be chosen for the new routine? (b) Find the probability that both experienced lifters end up in the same group. (c) What does that probability say about relying on random assignment alone, and what design change removes the worry? (d) The new routine group averages 4.2 kg more improvement, and the researcher writes that the new routine causes greater strength gains in adults. Critique that claim.

Experimental design: answer key

  1. (a) Units: the 36 seedlings; explanatory: lighting condition; response: height in cm after 4 weeks. (b) 3 treatments, 363=12\frac{36}{3} = 12 seedlings each. (c) Direct control of other sources of variation. (d) Yes, causation is supportable for these seedlings because the lighting was randomly assigned, but the result does not generalize beyond seedlings like these.

  2. (a) Explanatory: sport participation; response: GPA. (b) Observational study, because students self-selected and no treatment was assigned. (c) 603+933480=3.20\frac{603 + 933}{480} = 3.20, with a gap of 0.240.24 GPA points. (d) Time management and academic motivation: students who already have those habits are both more likely to commit to a team and more likely to earn a high GPA, so the two effects on GPA cannot be separated. (e) Random assignment of participation, which is not feasible here, so the paper should report association only.

  3. Number the 84 volunteers 1 to 84, generate random integers from 1 to 84 ignoring repeats until 42 distinct numbers appear, give those 42 the tablet and the other 42 an identical placebo, hold dosing time and instructions the same, hide the assignment from both the volunteers and the nurse recording scores, then compare the two groups' mean 0 to 24 symptom scores after 2 weeks. The placebo isolates the active ingredient from the experience of treatment; the double blinding keeps expectations out of the response and the measurement.

  4. (a) Single-blind, since only the patients are kept unaware. (b) The therapist's expectations can inflate the judgment-based scores he gives to patients he knows received the active cream. (c) Have a therapist with no access to the assignments do the scoring, making the study double-blind. (d) Without the inert cream, the ingredient and the experience of being treated would change together, so the placebo effect would be confounded with the ingredient.

  5. (a) Experience strongly affects finish time, so blocking on it removes that variation and sharpens the program comparison. (b) Two blocks, randomize inside each: 18 and 18 within the 36 experienced runners, 12 and 12 within the 24 novices. (c) 18+12=3018 + 12 = 30 runners per program. (d) No change in kind: random assignment within blocks still supports causation for these runners, and blocking improves precision only.

  6. (a) Matched pairs; the pairing controls each volunteer's own typing skill. (b) A coin flip per volunteer sets the order, giving 28=2562^8 = 256 equally likely order assignments across 8×2=168 \times 2 = 16 recorded speeds. (c) Differences 3,−1,5,5,2,4,3,33, -1, 5, 5, 2, 4, 3, 3, summing to 24, so dˉ=3.0\bar{d} = 3.0 wpm favoring layout B. (d) It keeps practice and fatigue from being confounded with layout.

  7. (a) 2×3=62 \times 3 = 6 treatments. (b) The two halves are blocks of 24 pots; 246=4\frac{24}{6} = 4 pots per treatment per block and 4×2=84 \times 2 = 8 pots per treatment overall, using all 48. (c) Number a half's 24 pots, draw 4 distinct random integers for each of five treatments ignoring repeats, give the 4 pots never drawn the sixth treatment, then repeat in the other half. (d) No. Every treatment appears 4 times in each half, so light varies with the block rather than with the treatments.

  8. (a) (2010)=184,756\binom{20}{10} = 184{,}756. (b) 919≈0.474\frac{9}{19} \approx 0.474, confirmed by counting 2(188)=87,5162\binom{18}{8} = 87{,}516 favorable assignments out of (2010)=184,756\binom{20}{10} = 184{,}756. (c) Random assignment balances on average, not in every single run, so block on experience (1 lifter to each routine, then a random 9 and 9 split of the other 18) to guarantee the balance. (d) If the 4.2 kg gap is larger than chance variation would produce, the causal part is defensible for these 20 volunteers because treatments were randomly assigned, but the leap to adults in general is not, since the volunteers were self-selected from one mailing list rather than randomly selected.

The key gives final answers only. The full worked solution for each problem is on the matching practice set, and every other sheet is on the worksheets page.

More worksheets from this unit