Choose your reading route
eduKateSG · Why Mathematics? · Latin Square Designs: Row and Column Blocking, Orthogonality and Missing Cells
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 8–14
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 15–21
Check assumptions
See where design and calibration matter.Route 4 · Chapters 22–28
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 29–42
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Why is mathematics important when two nuisance gradients affect an experiment at once? A Latin square places each treatment exactly once in every row and every column. This orthogonal balance controls two blocking factors with only k squared runs, provided additivity and randomisation are credible.
Did You Know? A 4×4 Latin square compares four treatments across four row levels and four column levels in 16 runs. A full three-factor 4×4×4 arrangement would require 64 runs, but the square cannot estimate all interactions.
Full chapter index
Chapters 1–7 · Understand the question
Chapters 8–14 · Reproduce the example
Chapters 15–21 · Check assumptions
Chapters 22–28 · Explore applications
Chapters 29–42 · Study and verify
- What students should transfer
- A spreadsheet workflow
- A coding workflow
- Applications and careers
- A student investigation
- Common misconceptions
- Connection to recent mathematics articles
- Guidance for students and parents
- Practice questions
- Frequently asked questions
- A reporting checklist
- The missing interactions are the price
- A final Latin-square audit
- Useful next reading
Rows and columns represent two nuisance factors; symbols represent the treatment of interest. Every role must be defined before outcomes are observed.
In a k×k Latin square, each of k treatments appears once in every row and once in every column.
Use rows ABCD, BCDA, CDAB and DABC. Every row and column contains A, B, C and D exactly once.
Treatment totals are balanced across both nuisance factors. Row and column main effects do not contaminate treatment contrasts under the additive model.
Four treatments with two four-level nuisance factors use 16 cells. The saving relative to 64 full combinations comes from not estimating interactions.
NIST explains the two blocking factors, row-column balance, additive model, run counts and assumptions. Read the NIST handbook.
| Row/Column | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| 1 | A | B | C | D |
| 2 | B | C | D | A |
| 3 | C | D | A | B |
| 4 | D | A | B | C |
Select a valid square, randomly relabel treatments, randomly permute rows and randomly permute columns. The printed cyclic form alone is not randomisation.
Use response=overall mean+row effect+column effect+treatment effect+error. Treatment-by-row and treatment-by-column interactions are assumed absent.
Suppose adjusted means are A=12, B=15, C=14 and D=19. D exceeds A by 7 under the additive model.
Because treatment is balanced across rows and columns, additive-model treatment means separate systematic nuisance shifts from treatment averages.
For k=4, treatment, row and column each use 3 degrees of freedom, leaving (k−1)(k−2)=6 error degrees of freedom.
With one observation per cell, interactions are confounded with residual variation. Strong row-treatment interaction can invalidate the simple conclusion.
Use major, crossed nuisance gradients such as day and machine or classroom and lesson period. Both need exactly k levels in the standard square.
| A | B | C | D |
|---|---|---|---|
| 12 | 15 | 14 | 19 |
One failed run breaks orthogonality. A model can estimate with reduced precision, but hand substitution should not pretend the observation existed.
Connectivity can fail depending on pattern. Inspect model rank and estimability before reporting treatment comparisons.
Replicate whole squares across sites or batches to estimate broader variation and examine treatment interactions with replication factors.
Superimpose a second orthogonal symbol set to block a third nuisance factor. Existence and orthogonality depend on the order.
A pair of orthogonal Latin squares does not exist for order 6, so a standard Graeco-Latin plan is impossible there.
Sequences can use Latin or Williams squares to balance treatment order and periods, though carryover requires additional properties.
Four practice methods are tested across four classes and four lesson periods. Each method appears once per class and once per period.
Treatments are arranged across field rows and columns to control two spatial fertility gradients.
Four settings are tested across four machines and four shifts. Machine-by-treatment interaction remains a risk.
Power depends on residual variance after row and column adjustment, number of treatments, replications and effect. One square can have limited error degrees of freedom.
Plot residuals by row, column and treatment, inspect spatial patterns, and test interaction only when replication makes it estimable.
Verify every symbol appears once per row and column, calculate treatment means and refit after deleting one cell.
Latin squares require equal numbers of levels, additive effects, complete execution and credible randomisation. They are not universal small-sample shortcuts.
Publish the randomised square, factor meanings, run order, missing cells, model, treatment estimates, row and column effects and residual checks.
Use a Latin square when two strong crossed nuisance factors have the same number of levels as treatments and interaction is plausibly small.
Orthogonality is planned balance. It lets one comparison stay clear of other structured variation, but only for effects the design represents.
Represent the square as tidy rows with row block, column block, treatment, randomised order, response and missing flag.
Verify symbol counts, matrix rank and orthogonality; fit additive effects and test missing-cell estimability.
Agronomists, manufacturers, sensory scientists and education researchers use Latin squares when two nuisance gradients must be controlled economically.
Arrange four paper types across four days and four operators. Randomise labels and order, then simulate a row-treatment interaction.
A Latin square is not merely a puzzle, balance does not estimate interactions, and a cyclic square is not random until labels and order are randomised.
The randomised-block guide explains one nuisance factor. The factorial guide explains what the square sacrifices.
Students should check rows and columns manually before analysis. Parents can ask whether a school experiment confused a teaching method with class or lesson time.
- Verify a 4×4 square.
- Randomise labels.
- Calculate treatment means.
- Delete one cell.
- Add a treatment-row interaction.
### Why “Latin”? Treatments are traditionally shown as letters.
### How many nuisance factors? Two in a standard Latin square.
### Can it estimate interactions? Not in the basic unreplicated form.
### Must levels match? Yes, in the standard design.
Report treatments, row and column factors, selected square, randomisation, run order, outcomes, missing cells, additive model, treatment estimates, residual degrees of freedom and diagnostics.
Sixteen runs replace sixty-four combinations because the design assumes nuisance interactions are negligible.
Confirm row and column balance survived execution, randomisation was real and every reported contrast is estimable under the actual data pattern.
Continue through the Mathematics Learning Hub. Recalculate the central example, identify the assumptions and explain which claim the result can and cannot support.
Extended practice laboratory
Practice case 1: Choose treatments
Setup: Start from the article’s Latin square mechanism for case 1. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 2: Choose rows
Setup: Start from the article’s Latin square mechanism for case 2. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 3: Choose columns
Setup: Start from the article’s Latin square mechanism for case 3. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 4: Check equal levels
Setup: Start from the article’s Latin square mechanism for case 4. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 5: Create base square
Setup: Start from the article’s Latin square mechanism for case 5. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 6: Randomise treatment labels
Setup: Start from the article’s Latin square mechanism for case 6. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 7: Randomise rows
Setup: Start from the article’s Latin square mechanism for case 7. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 8: Randomise columns
Setup: Start from the article’s Latin square mechanism for case 8. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 9: Verify row balance
Setup: Start from the article’s Latin square mechanism for case 9. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 10: Verify column balance
Setup: Start from the article’s Latin square mechanism for case 10. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 11: Record outcomes
Setup: Start from the article’s Latin square mechanism for case 11. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 12: Calculate treatment means
Setup: Start from the article’s Latin square mechanism for case 12. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 13: Fit additive model
Setup: Start from the article’s Latin square mechanism for case 13. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 14: Estimate row effects
Setup: Start from the article’s Latin square mechanism for case 14. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 15: Estimate column effects
Setup: Start from the article’s Latin square mechanism for case 15. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 16: Estimate treatment effects
Setup: Start from the article’s Latin square mechanism for case 16. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 17: Check interactions
Setup: Start from the article’s Latin square mechanism for case 17. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 18: Add missing cell
Setup: Start from the article’s Latin square mechanism for case 18. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 19: Recover with model
Setup: Start from the article’s Latin square mechanism for case 19. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 20: Check residual df
Setup: Start from the article’s Latin square mechanism for case 20. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 21: Compare block design
Setup: Start from the article’s Latin square mechanism for case 21. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 22: Add Graeco-Latin factor
Setup: Start from the article’s Latin square mechanism for case 22. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 23: Simulate nuisance gradients
Setup: Start from the article’s Latin square mechanism for case 23. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 24: Name limits
Setup: Start from the article’s Latin square mechanism for case 24. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 25: Write conclusion
Setup: Start from the article’s Latin square mechanism for case 25. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.