In m-harikiran/T-Test: Hypotheisis testing for two samples

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Introduction

The package ADVTTEST can be used to do Hypothesis Testing using a T-Test to decide if the population means of two samples are identical or not. This package will run a T-Test on paired samples, unpaired samples, and samples with equivalent and unequal variances.

Package Installation

The below are the steps to install the package ADVTTEST

1. Install project R

2. Install R Studio

3. Open R Studio and install devtools using command install.packages("devtools")

4. Install the package ADVTTEST using command devtools::install_github('https://github.com/m-harikiran/T-Test.git', auth_token = <token>, build_vignettes = TRUE)

Package Functions

This package as the below functions:

1. myttest a constructor function

2. plot.Rttest a generic plot function for class Rttest

3. print.Rttest a generic print function for class Rttest

4. shinyttest a function used to run reactive shiny web application

Constructor

The constructor function is called myttest, and it takes four input samples as vectors, a logical variable (default FALSE) and a numeric variable (default 0.05). Performs hypothesis testing, returning a list of class Rttest objects containing the test type, test summary, hypothesis test conclusion, and input sample data.

# Checking if the given input vectors are paired or not
if (paired == FALSE)
{
  #If the given samples are not paired then
  #Perform F-Test to know the variability of the samples

  ftest <- var.test(x, y)
  #alternative hypothesis: true ratio of variances is not equal to 1
  #If p-value is > alpha or significance level we fail to reject null

  if (ftest$p.value > alpha) {
    #Equal variance
    testType <- 'T-TEST'

    ttest <-
      t.test(x, y, var.equal = TRUE, conf.level = 1 - alpha)

    if (ttest$p.value > alpha) {
      conclusion <- 'N'
    } else {
      conclusion <- 'Y'
    }
  }
  else {
    #Unequal Variance
    testType <- 'WELCH'

    ttest <-
      t.test(x, y, var.equal = FALSE, conf.level = 1 - alpha)

    if (ttest$p.value > alpha) {
      conclusion <- 'N'
    } else {
      conclusion <- 'Y'
    }
  }

}
else{
  #Given samples are paired
  testType <- 'PAIRED'

  if (length(x) != length(y))
    stop('For Paired Samples Length of X and Y must be same')

  ttest <- t.test(x, y, paired = TRUE, conf.level = 1 - alpha)

  if (ttest$p.value > alpha) {
    conclusion <- 'N'
  } else {
    conclusion <- 'Y'
  }
}

#List containing the results
lst = list(
  'Test_Type' = testType,
  'Test_Conclusion' = conclusion,
  'Test_Summary' = ttest,
  'Data' = inputDF
)

Print Generic

The method print.Rttest is a generic print method for class Rttest. This method takes a class object as an input and prints confidence interval with confidence level and Y or N conclusion to reject NULL hypothesis.

print.Rttest <- function(x, ...) {
  if (!is.list(x))
    stop('The input is not a list')
  if (class(x) != 'Rttest')
    stop('The given input is not an object of class Rttest')

  ci <-
    x$Test_Summary$conf.int   #Accessing the confidence interval from test summary

  type <- x$Test_Type

  lst = list('Confidence Interval' = ci, 'Test Type' = type)

  print(lst)
}

Plot Generic

The method plot.Rttest is a generic plot method for class Rttest. This method takes a class object as an input and plots box plot for unpaired data and box plot of differences if data is paired .

#Box Plot for Non-Paired samples
df = x$Data
gplot <-
  ggplot(df,
         aes(x = Variable,
             y = Data,
             fill = Variable)) + geom_boxplot(outlier.colour = "blue",
                                              outlier.size = 2) + labs(title = "Box Plot for unpaired samples",
                                                                       x = "Sample",
                                                                       y = "Values",
                                                                       fill = "Sample") + theme_bw()

#Box Plot for PAIRED Samples
df <-
  data.frame('data' = xValues - yValues, Variable = 'x') #Difference between x and y

gplot <-
  ggplot(df,
         aes(x = Variable,
             y = data)) + geom_boxplot(outlier.color = 'red', outlier.size = 2) + labs(title = "Box Plot for difference between paired samples", x = "Difference", y = "Values(x-y)") +   geom_errorbar(aes(
               ymin  = x$Test_Summary$conf.int[1],
               ymax  = x$Test_Summary$conf.int[2],
               width = 0.01
             )) + theme_bw()

Shinyttest Method

The method shinyttest is used to run shiny web application, which can be used to interact with ADVTTEST package in an interactive way. This application displays the input samples data, summary of Hypothesis Testing and box plot's for paired and unpaired samples.

#The code to run the shiny application
shiny::runApp(system.file("ttest", package = "ADVTTEST"), launch.browser = TRUE)

Shiny Application

This is a web application which can be used to interact with ADVTTEST package in an interactive way. The application can display the samples data, test statistics and box plots in a reactive way. The shiny application has two main components or functions UI and SERVER. In ui method all the components related to ui are defined and in server method all the back end functions are defined. The application can be opened using the command ADVTTEST::shinyttest()

Testing Package

library(ADVTTEST)

Non Paired Salples with Eqlal Variances

set.seed(32); x=rnorm(30,mean=10,sd=15)
set.seed(35); y=rnorm(30,mean=8,sd=15)
ans1=ADVTTEST::myttest(x,y,alpha=0.05,paired=FALSE)

print(ans1)

plot(ans1)

Non Paired Salples with Uneqlal Variances

set.seed(32); x=rnorm(30,mean=10,sd=5)
set.seed(35); y=rnorm(30,mean=8,sd=15)
ans2=ADVTTEST::myttest(x,y,alpha=0.05,paired=FALSE)

print(ans2)

plot(ans2)

Paired Salples

set.seed(32); x=rnorm(30,mean=10,sd=15)
set.seed(35); y = x+ rnorm(30, 5 ,4)
ans3=ADVTTEST::myttest(x,y,alpha=0.05,paired=TRUE)

print(ans3)

plot(ans3)

m-harikiran/T-Test documentation built on Dec. 21, 2021, 12:47 p.m.