Statistics between 3 sets of genes

Question

[anu014]

Hello Biostars!

Today I am dealing with Statistics. I have 3 sets of genes (from same organism) with their AT% content. Now, I want to compare these sets as to which sets are more different from each other.

I tried Type-III ANOVA. But I'm not sure it's the correct method to apply :

library("car")

df1 <- data.frame(file1$at, file1$sample)  #sample is gene_set1
#eg df1
#       at                       sample
#1 42.5852 gene_set1
#2 41.8838 gene_set1

df2 <- data.frame(file2$at, file2$sample) #sample is gene_set2
#eg df2
#       at                       sample
#1 40.5852 gene_set2
#2 45.8838 gene_set2

df3 <- data.frame(file3$at, file3$sample) #sample is gene_set3
#eg df3
#       at                       sample
#1 40.5852 gene_set3
#2 41.3458 gene_set3

up_unres <- rbind(df1, df2)
down_up <- rbind(df3, df1)
down_unres <- rbind(df3, df2)

Anova(lm(at ~ sample, data=up_unres), type="III")
Anova(lm(at ~ sample, data=down_up), type="III")
Anova(lm(at ~ sample, data=down_unres), type="III")

Can anyone tell me which statistical method will be precise enough for this?

Please do suggest. Thank you :)

[russhh]

you only have one factor, why did you choose to use type-III anova?

[anu014]

Because my data is unbalanced. The number of genes in each set is different.

[russhh]

but balance is only really a problem for factorials. Do a single anova over your three groups anova(lm(at ~ sample, data = my_data)), R will deal with the unequal set sizes.

Answer 1

[Nicolas Rosewick]

You shoud use anova followed by Tukey's test.

Check here : https://www.r-bloggers.com/anova-and-tukeys-test-on-r/

Something like this (not tested) :

df <- rbind(df1,df2,df3)
res.aov <- anova(lm(at ~ sample, data=df))

res.tukey <- TuckeyHSD(x=aov(at ~ sample,data=df),'sample',conf.level=0.95)