Friedman-like test for a two-way layout with an equal number
  (cc>1) of replications per cell
See Section 7.9 in the text (Hollander & Wolfe 2nd edn)

Test for consistency across 4 Laboratories for amounts of niacin
  detected in enriched bran flakes
Data Source: Campbell and Pelletier (1962)
  (See Table 7.20 in text, page 331.)
k=4 treatment groups, n=3 blocks, and c=3 replications
  per (treatment,block) combination
Total number of observations:  4*3*3 = 36

Starting random-number seed: 12345
(Enter a number on the command line to specify a different seed.)
Using 100,000 permutations in permutation tests.

       Low Enrichment      Med Enrichment      High Enrichment     
Lab1:  7.58  7.87  7.71   11.63 11.87 11.40   15.00 15.92 15.58  (Av=11.62)
Lab2:  8.00  8.27  8.00   12.20 11.70 11.80   16.60 16.40 15.90  (Av=12.10)
Lab3:  7.60  7.30  7.82   11.04 11.50 11.49   15.87 15.91 16.28  (Av=11.65)
Lab4:  8.03  7.35  7.66   11.50 10.10 11.70   15.10 14.80 15.70  (Av=11.33)

Midranks within blocks (1 le Midrank le 12):
       Low Enrichment      Med Enrichment      High Enrichment     
Lab1:   3     8     6       7    11     3       2     9     4    (AvR= 5.89)
Lab2:   9.5  12     9.5    12     8.5  10      12    11     7    (AvR=10.17)
Lab3:   4     1     7       2     5.5   4       6     8    10    (AvR= 5.28)
Lab4:  11     2     5       5.5   1     8.5     3     1     5    (AvR= 4.67)

Mack-Skilling test statistic  MS=12.927
  Large-sample chi-square approx P-value:  P=0.00480  (df=3)

Rank sums (for the 4 treatments):    53    91.5    47.5    42
Permutation score = Sum of squares of treatment-group
  rank sums = 15201.5
Carrying out 100,000 Friedman-like permutations to estimate
  the exact P-value for permutation score 15201.5:
  P = 236/100000 = 0.00236  95% CI (0.0021, 0.0024, 0.0027)

Multiple comparisons based on Mack-Skillings S differences:
Approximate multiple-comparison-corrected P-values for treatment
  differences based on the normal-range statistic for k=4 normals
  (based on 100,000 simulations):
Lab1 vs. Lab2:  Sdiff=12.83   P =  5613/100,000 = 0.05613 
Lab1 vs. Lab3:  Sdiff= 1.83   P = 98367/100,000 = 0.98367 
Lab1 vs. Lab4:  Sdiff= 3.67   P = 88812/100,000 = 0.88812 
Lab2 vs. Lab3:  Sdiff=14.67   P =  2046/100,000 = 0.02046 (*)
Lab2 vs. Lab4:  Sdiff=16.50   P =   670/100,000 = 0.00670 (**)
Lab3 vs. Lab4:  Sdiff= 1.83   P = 98367/100,000 = 0.98367 

Using 100,000 sets of within-block permutations to simulate the exact
  multiple-comparison-corrected P-values for all pairwise differences:
Lab #1 vs. #2:  Sdiff=12.83   P =  1065/100,000 = 0.01065 (*)
Lab #1 vs. #3:  Sdiff= 1.83   P = 73547/100,000 = 0.73547 
Lab #1 vs. #4:  Sdiff= 3.67   P = 48552/100,000 = 0.48552 
Lab #2 vs. #3:  Sdiff=14.67   P =   314/100,000 = 0.00314 (**)
Lab #2 vs. #4:  Sdiff=16.50   P =    87/100,000 = 0.00087 (**)
Lab #3 vs. #4:  Sdiff= 1.83   P = 73524/100,000 = 0.73524 

Random numbers used:  7,109,110