applying two-sample t.test comparing multiple groups in two categories

Question

I want to do a two-sample t.test comparing values from two regions for several groups. I know it must be a simple trick, but I couldn't find an answer so far.

This is a short version of my table. Several species each from 2 regions (There are generally 6 specimens per species available for each region).

species region   N15     
ARGAFF     EQ    9.85     
ARGAFF     EQ   10.42   
ARGAFF     EQ   10.43  
ARGAFF    OMZ   10.28  
ARGAFF    OMZ   10.30   
ARGAFF    OMZ   10.41  
BATABY     EQ   10.57      
BATABY     EQ   10.60      
BATABY     EQ   10.68        
BATABY    OMZ    9.21       
BATABY    OMZ    9.29       
BATABY    OMZ    9.67       

I tried

spp<-split(SIA,species,drop=FALSE) 

followed by

lapply(spp,function (x) t.test(N15~region))

I further tried

by(SIA[,3], SIA[, "species"], function (x) t.test(N15~region))

but both versions compute the test for all values (N.138, not just for individual species (N:12).

On the other hand

 by(SIA[,3], SIA[, "species"], t.test)

results in a one-sample test but for individual species.

I'd appreciate advice.

Answer 1

There are various ways to do that, but I'm using a combination of dplyr and broom packages. The advantage of this process is that you export the t-test information as a dataset:

library(dplyr)
library(broom)

set.seed(9)

# example dataset
dt = data.frame(species = c(rep("ARGAFF",6), rep("BATABY",6)),
                region = rep(c("EQ","OMZ"),6),
                N15 = rnorm(12,10,1))


dt_result = dt %>% group_by(species) %>% do(tidy(t.test(N15~region, data=.)))

dt_result

   #   species  estimate estimate1 estimate2 statistic   p.value parameter   conf.low conf.high
   # 1  ARGAFF 0.6029705  9.842659  9.239688  1.381967 0.2439265  3.732157 -0.6434982  1.849439
   # 2  BATABY 1.0238324 10.740491  9.716659  1.994738 0.1673019  2.300604 -0.9298488  2.977514

Hope it works for you.

Answer 2

First, in general it is a bad idea to attach a dataset. You can modify your lapply statement to do what you want I think.

set.seed(9)

# example dataset
SIA <- data.frame(species = as.factor(c(rep("ARGAFF",6), rep("BATABY",6))),
                region = as.factor(rep(c("EQ","OMZ"),6)),
                N15 = rnorm(12,10,1))

spp <- split(SIA, SIA$species,drop=FALSE) 
out <- vector("list", length = length(spp))
out <- lapply(1:length(spp),function (x) out[[x]]<- t.test(spp[[x]]$N15 ~ spp[[x]]$region))

Then you can extract whatever features you want.

> out
[[1]]

    Welch Two Sample t-test

data:  spp[[x]]$N15 by spp[[x]]$region
t = -0.7566, df = 2.4628, p-value = 0.515
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 -4.617903  3.018924
sample estimates:
 mean in group EQ mean in group OMZ 
         9.982135         10.781625 


[[2]]

    Welch Two Sample t-test

data:  spp[[x]]$N15 by spp[[x]]$region
t = 0.66769, df = 3.2858, p-value = 0.5483
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 -1.564611  2.448523
sample estimates:
 mean in group EQ mean in group OMZ 
        10.055578          9.613622 

Answer 3

Rstatix::t_test() is a user-friendly wrapper of t.test() that returns a tibble.

library(tidyverse)
library(rstatix)

df <- tribble(~species, ~region,  ~N15,     
              "ARGAFF",    "EQ",  9.85,     
              "ARGAFF",    "EQ", 10.42,   
              "ARGAFF",    "EQ", 10.43,  
              "ARGAFF",   "OMZ", 10.28,  
              "ARGAFF",   "OMZ", 10.30,   
              "ARGAFF",   "OMZ", 10.41,  
              "BATABY",    "EQ", 10.57,      
              "BATABY",    "EQ", 10.60,      
              "BATABY",    "EQ", 10.68,        
              "BATABY",   "OMZ",  9.21,       
              "BATABY",   "OMZ",  9.29,       
              "BATABY",   "OMZ",  9.67)
df %>%
  group_by(species) %>%
  t_test(N15~region)