How to run the Kruskal-Wallis or Mann-Whitney Test in R?

Question

Can anyone give me a hint on how to run the Kruskal-Wallis Test below?

My objective : Is there any significance of the growth (agg_rel_abund) of bacteria between Forest and Urban for each family.

The code I have tried in R : kruskal.test(Habitat ~ agg_rel_abund, data = my_data) but obviously I know that is wrong... because I didn't hit my objective..

Let me briefly explain about my data :

There are types of sample, which is F and W.

When the sample name start with F, it means the Habitat is from Urban.

When the sample name start with W, it means the Habitat is from Forest.

It is okay if want to perform Mann-Whitey Test, or any Non-Parametric Test too... as long as can get to know the significance of the growth (agg_rel_abund) of bacteria between Forest and Urban for each family.

Sample	Habitat	Family	agg_rel_abund
F10	Urban	Acetobacteraceae	0
F2	Urban	Acetobacteraceae	0
F3	Urban	Acetobacteraceae	0
F7	Urban	Acetobacteraceae	0.000132118
F8	Urban	Acetobacteraceae	0
W10	Forest	Acetobacteraceae	0
W13	Forest	Acetobacteraceae	0
W3	Forest	Acetobacteraceae	0
W6	Forest	Acetobacteraceae	0
W9	Forest	Acetobacteraceae	0
F10	Urban	Bacillaceae	0.00488836
F2	Urban	Bacillaceae	0.000924825
F3	Urban	Bacillaceae	0.001056943
F7	Urban	Bacillaceae	0.002378121
F8	Urban	Bacillaceae	0.002906593
W10	Forest	Bacillaceae	0.000264236
W13	Forest	Bacillaceae	0.027876866
W3	Forest	Bacillaceae	0.001585414
W6	Forest	Bacillaceae	0.001056943
W9	Forest	Bacillaceae	0.004492007
F10	Urban	Carnobacteriaceae	0
F2	Urban	Carnobacteriaceae	0
F3	Urban	Carnobacteriaceae	0
F7	Urban	Carnobacteriaceae	0
F8	Urban	Carnobacteriaceae	0.000132118
W10	Forest	Carnobacteriaceae	0
W13	Forest	Carnobacteriaceae	0
W3	Forest	Carnobacteriaceae	0.000132118
W6	Forest	Carnobacteriaceae	0

kruskal.test(agg_rel_abund ~ Habitat, data = my_data) The dependent variable should be placed before ~. It should be like, Dep.var ~Independent.var — Mohanasundaram, Jul 08 '21 at 13:50
@Mohanasundaram , I don't need to run the test for every type of bacteria to get the p-value right? Just wondering... I should just use kruskal.test(agg_rel_abund ~ Habitat, data = my_data) right? — Gambit, Jul 08 '21 at 13:58

Mohanasundaram · Accepted Answer · 2021-07-08T17:39:31.017

This question should be in cross-validated.

If you want to know whether the the growth is varying with Family, irrespective of the Habitat, you can perform kruskal.test with agg_rel_abund as dependent variable and Family as independent variable.

kruskal.test(agg_rel_abund ~ Habitat, data = my_data)

Kruskal-Wallis rank sum test

data:  agg_rel_abund by Habitat
Kruskal-Wallis chi-squared = 0.0051556, df = 1, p-value = 0.9428

If you are sure that there is no difference in growth across different families, you can directly perform kruskal.test with agg_rel_abund as dependent variable and Habitat as independent variable.

kruskal.test(agg_rel_abund ~ Habitat, data = my_data)

Kruskal-Wallis rank sum test

data:  agg_rel_abund by Habitat
Kruskal-Wallis chi-squared = 0.0051556, df = 1, p-value = 0.9428

For each habitat, you can perform kruskal.test to check the significant of difference in growth among families

library(dplyr)

    for (i in unique(family$Habitat)) {
  x <- kruskal.test(agg_rel_abund ~ family,
                    data = family[family$Habitat==i,])
  out[[i]] <- c(Kruskal.Wallis.H = x[["statistic"]][["Kruskal-Wallis chi-squared"]],
                Sig = x[["p.value"]],
                df = x[["parameter"]][["df"]])
  }

out <- bind_rows(out)
out$Habitat <- unique(family$Habitat)

How to run the Kruskal-Wallis or Mann-Whitney Test in R?

1 Answers1