kruskalmc {pgirmess}

Multiple comparison test after Kruskal-Wallis

Description

Multiple comparison test between treatments or treatments versus control after Kruskal-Wallis test

Usage

kruskalmc(resp,...)
## Default S3 method:
kruskalmc(resp, categ, alpha = 0.05, cont=NULL,...)
## S3 method for class 'formula'
kruskalmc(resp,data=NULL,...)

Arguments

resp

a numeric vector of data values or a formula of the type 'response~category'.

categ

a factor object giving the group for the corresponding elements of 'x'

alpha

the significance level

cont

NULL (default) for multiple comparison between treatments; 'one-tailed' or 'two-tailed' for corresponding multiple comparisons treatments versus control; partial matching allowed

data

a data.frame including the variables used in the formula

...

other parameters to be passed as arguments (not used here)

Details

When the value of a Kruskal-Wallis test is significant, it indicates that at least one of the groups is different from at least one of the others. This test helps determining which groups are different with pairwise comparisons adjusted appropriately for multiple comparisons. Those pairs of groups which have observed differences larger than a critical value are considered statistically different at a given significance level. Three types of multiple comparisons are implemented: comparisons between treatments, 'one-tailed' and 'two-tailed' comparison treatments versus control. The first factor level is considered the control. NAs are omitted from data before processing.

For further details please consider the refence below where the method is fully described, or visit https://giraudoux.pagesperso-orange.fr/#pgirmess where a copy of the corresponding book section can be downloaded.

Value

A list of class 'mc' with the following items:

statistic

statistics used

signif.level

the significance level

dif.com

a data.frame with observed and critical differences

Note

Alternative methods are proposed in the section 'see also', on François Gillet's suggestion. The three methods do not give necessarily the same results, and the why is still to investigate

References

Siegel and Castellan (1988) Non parametric statistics for the behavioural sciences. MacGraw Hill Int., New York. pp 213-214

See Also

kruskal.test; to reorder factor levels see relevel; for other functions about median multiple comparison see package 'PMCMRplus'; kruskal

Examples

resp<-c(0.44,0.44,0.54,0.32,0.21,0.28,0.7,0.77,0.48,0.64,0.71,0.75,0.8,0.76,0.34,0.80,0.73,0.8)
categ<-as.factor(rep(c("A","B","C"),times=1,each=6))
kruskalmc(resp, categ)
kruskalmc(resp, categ, alpha=0.01)
kruskalmc(resp, categ, cont="one-tailed")
kruskalmc(resp, categ, cont="two-tailed")

kruskalmc(resp~categ)
kruskalmc(resp~categ, alpha=0.01)
kruskalmc(resp~categ, cont="one-tailed")
kruskalmc(resp~categ, cont="two-tailed")