Nonparametric Tests

Friedman Test Table: Exact Critical Values

Exact critical values for the Friedman test, the nonparametric alternative to repeated-measures ANOVA.

  • k = 3 to 6 treatments
  • Exact values
  • Exact calculator

Friedman Test Table

Jump to a value of k below. Rows: number of blocks (subjects) · Columns: α · Cells: critical Friedman statistic. Reject H₀ if the statistic ≥ the cell.

k = 3 treatments

k = 3 treatments
α →
n (blocks) ↓
0.10.050.0250.010.0050.001
36.0006.000————
46.0006.5008.0008.0008.000—
55.2006.4007.6008.40010.00010.000
65.3347.0008.3349.00010.33412.000
75.4297.1437.7158.85810.28612.286
85.2506.2507.7509.0009.75012.250
95.5566.2238.0009.55610.66712.667
105.0006.2007.8009.60010.40012.600
115.0916.5467.8199.45510.36413.273
125.1676.5008.0009.50010.50012.667
134.7706.6167.5399.3859.84712.462
145.1436.1437.4299.14310.42913.286
154.9346.4007.6008.93410.53412.934
164.8756.5007.6259.37510.12513.500
175.0596.1187.4129.29510.70613.059
184.7786.3347.4459.00010.33413.000
195.0536.4227.6859.57910.21113.369
204.9006.3007.5009.30010.80013.300

k = 4 treatments

k = 4 treatments
α →
n (blocks) ↓
0.10.050.0250.010.0050.001
26.0006.000————
36.6007.4008.2009.0009.000—
46.3007.8008.4009.60010.20011.100
56.3607.8008.7609.96010.92012.600
66.4007.6008.80010.20011.40012.800
76.4297.8009.00010.54311.40013.458
86.3007.6509.00010.50011.55013.800
96.2007.6678.86710.73411.80014.067
106.3607.6809.00010.68011.88014.520
116.2737.6919.00010.74612.05514.564
126.3007.7009.10010.80012.10014.800
136.1397.8009.09310.84712.23114.908
146.3437.7159.08610.88612.25815.086
156.2807.7209.16010.92012.20015.080

k = 5 treatments

k = 5 treatments
α →
n (blocks) ↓
0.10.050.0250.010.0050.001
27.2007.6008.0008.000——
37.4678.5349.60010.13410.66711.467
47.6008.8009.80011.20012.00013.200
57.6808.96010.24011.68012.48014.400
67.7349.06710.40011.86713.06715.200
77.7729.14310.51512.11513.25815.658
87.7009.20010.60012.30013.50016.000
97.7349.24510.66712.44513.68916.356
107.7609.28010.72012.48013.84016.560

k = 6 treatments

k = 6 treatments
α →
n (blocks) ↓
0.10.050.0250.010.0050.001
28.2869.1439.4299.71510.000—
38.7159.85810.81011.76212.52413.286
49.00010.28611.42912.71513.57215.286
59.00010.48611.74313.22914.25816.429
69.04810.57212.00013.62014.76217.048

Tip: click any value to highlight its row and column.

How to read this table

  1. Rank the k treatments within each block (subject) and add up the ranks Rⱼ for each treatment over the n blocks.
  2. Compute Q = 12/(nk(k + 1)) Σ Rⱼ² − 3n(k + 1).
  3. Find the table for k and the row for n.
  4. Reject H₀ if Q is greater than or equal to the table value. Beyond the table, compare Q with chi-square on k − 1 df.

Worked example

Eight patients each try three formulations and rank them, giving Q = 6.75. In the k = 3 table, row 8, α = 0.05 gives 6.250, so the formulations differ at the 5% level.

Click the underlined links to highlight the value in the table.

Calculator

For values between the rows and columns of the table. Results update as you type.

Formula and how the values were computed

Q = 12 / (nk(k + 1)) · Σ Rⱼ² − 3n(k + 1)

Values come from the exact permutation distribution of Q, built block by block over all k! rank orderings. Each entry is the smallest attainable Q with P(Q ≥ q) ≤ α, rounded up. The table assumes no ties within blocks.

Frequently asked questions

What is a block?

A block is a set of related observations, one per treatment: usually one subject measured under every condition, as in a crossover trial.

Friedman or Page's L?

Friedman's test detects any difference among treatments. When you expect a specific order, such as increasing dose, Page's L test is more powerful.

Related calculators and tutorials