| T O P I C R E V I E W |
| Patch_Master |
Posted - 05/06/2011 : 4:10:38 PM
Hello,
Does anybody know how to run a post-hoc test in Origin Pro 8.1 after running a Kruskal Wallis ANOVA. |
| 3 L A T E S T R E P L I E S (Newest First) |
| Echo_Chu |
Posted - 05/09/2011 : 05:21:53 AM
For example, we have 3 groups as below.
group1 group2 group3
77 83 78
79 75 79
85 87 83
78 91 71
72 81 72
76 89 77
First we use Mann-Whitney 3 times to get p-value for comparison of each pair of groups
group1 v.s.group2 0.0927
group2 v.s.group3 0.0524
group1 v.s.group3 0.80887
Then have the 3 p-values to compare with 0.05/3=0.017. All the there p-values are larger than 0.017, that means each pair of three group is not significant different at 0.05 level.
|
| sherylradcliff |
Posted - 05/09/2011 : 04:19:32 AM
quote:
Originally posted by Echo_Chu
Hi, Patch
Origin does not support a direct post-hoc test for non-parametric analysis. However, you can do a relevant paried test for each pair with Mann-Whitney test and adjust the p value with Bonferroni correction. For example, if you have four groups, and you want the overall test of significance between them come out significant(at the 0.05 level), then the adjusted p value for each paired comparison is 0.05/4 = 0.0125. So any pair has to achieve a significant value on a paired test smaller than 0.0125 to be significant at the 0.05 level.
Echo
quote:
Originally posted by Patch_Master
Hello,
Does anybody know how to run a post-hoc test in Origin Pro 8.1 after running a Kruskal Wallis ANOVA.
can you please elaborate your answer more? thanks. |
| Echo_Chu |
Posted - 05/09/2011 : 01:55:53 AM
Hi, Patch
Origin does not support a direct post-hoc test for non-parametric analysis. However, you can do a relevant paried test for each pair with Mann-Whitney test and adjust the p value with Bonferroni correction. For example, if you have four groups, and you want the overall test of significance between them come out significant(at the 0.05 level), then the adjusted p value for each paired comparison is 0.05/4 = 0.0125. So any pair has to achieve a significant value on a paired test smaller than 0.0125 to be significant at the 0.05 level.
Echo
quote:
Originally posted by Patch_Master
Hello,
Does anybody know how to run a post-hoc test in Origin Pro 8.1 after running a Kruskal Wallis ANOVA.
|
|
|
