This preview shows page 1. Sign up to view the full content.
Unformatted text preview: ormative than the rank sum. This is obtained by dividing the
rank sum Ri for a given group by the number of companies in that group and is
shown in the table below for each group. If the null hypothesis is true, then these
three average ranks should all be relatively close together. In this case the average
ranks differ, this indicates the alternative hypothesis is true.
1<=DA<=5 6<=DA<=10 11<=DA<=14 10 5 4 9 12 2 13 6 7 11 3 1 14
t1= 5, R1=57, R/t = 11.4 8
t2=5, R2=34, R/t=6.8 t3=4, R3=14, R/t=3.5 3. The test statistic needs to reflect how different the average ranks are, that is it
should be a measure of the dispersion of the ranks. Leach provides the proof of
the formula (1979 pp149150), here it will be take as given: 211 K= 3(n+1)+ 12 Σ Ri2
n(n+1) ti
4. In this case, K=8.1428
There are 2 degrees of freedom in this case.
It is therefore possible to reject the null hypothesis at α=0.05
5. To find the locus of the difference, use the rank sum test and choose α=0.02 as the
significance level.
6. Writing the groups in order of increasing R/T:
11<=DA<=14 6<=DA<=10 1<=DA<=5 4 5 10 2 12 9 7 6 13 1 3 11 t1= 5, R1=57, R/t = 11.4 8
t2=5, R2=34, R/t=6.8 14
t3=4, R3=14, R/t=3.5 7. Then comparing 11<=DA<=14 with 1<=DA<=5 gives:
p’s q’s 5 0 5 0 5 0 5
P=20 0
Q=0 S=PQ=20, which is significant at α=0.02
8. Then comparing 11<=DA<=14 with 6<=DA<=10 gives: 212 p’s q’s 4 1 5 0 3 2 5
P=17 0
Q=3 S=PQ=14, which is not significant at α=0.02.
9. So the result of the Wilcoxon Rank Sum for PR is, 11<=DA<=14, 6<=DA<=10,
1<=DA<=5 213 APPENDIX 5:CRITICAL VALUES OF ρ FOR SPEARMAN TESTS Examples: For a two tailed test, with n=15 and α=0.05, reject the null hypothesis if
the obtained ρ is larger than or equal to 0.521, or if it is smaller than or equal to
0.521.
Examples: For a one tailed test, with n=15 and α=0.05, if an upper tail test is required,
reject the null hypothesis if the obtained is larger than or equal to 0.446. If a lower
tail test is required, reject if the obtained ρ is smaller than or equal to 0.446.
One tailed significance level, α
0.1 0.05 0.025 0.01 0.005 0.001 Two tailed significance level, α
n 0.2 0.1 0.05 0.02 0.01 0.002 4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 1
0.8
0.657
0.571
0.524
0.483
0.455
0.427
0.406
0.385
0.367
0.354
0.341
0.328
0.317
0.309
0.299 1
0.9
0.829
0.714
0.643
0.6
0.564
0.536
0.503
0.484
0.464
0.446
0.429
0.414
0.401
0.391
0.38 1
0.866
0.786
0.738
0.7
0.648
0.618
0.587
0.560
0.538
0.521
0.503
0.488
0.474
0.46
0.447 1
0.943
0.893
0.833
0.783
0.745
0.709
0.678
0.648
0.626
0.604
0.585
0.566
0.550
0.535
0.522 1.00
0.929
0.881
0.833
0.794
0.755
0.727
0.703
0.679
0.657
0.635
0.618
0.6
0.584
0.57 1.00
0.952
0.917
0.879
0.845
0.818
0.791
0.771
0.750
0.729
0.711
0.692
0.674
0.66 From Leach (1979) 214 APPENDIX 6:CRITICAL VALUES OF K FOR KRUSKAL WALLIS TEST WITH 3 INDEPENDENT
SAMPLES t1 is the number of observations in the largest sample
t3 is t...
View
Full
Document
 Summer '14
 The Land

Click to edit the document details