Improving_MultComp

This repository aims to improve the output generated from multiple comparisons

It is based on the paper: I know what I shouldn’t say, but what should I say? An approach to present results without statistical significance I really liked the table presented in this paper. It might be something interesting to consider in future analysis https://doi.org/10.3389/fams.2024.1487750

Because of that I want to create a new funcion to decribe the results in line with their suggestions

Tutorial to run from github

First download the folder with the scripts and then use the source function to extract the code

system("git clone https://github.com/Leitemfa/Improving_MultComp.git")

I want to make it in a way that you can extract their output from the simple lm model and also for multiple comparisons. For that last one, I made a code from the multcomp package.

Loading libraries

For running this code you will need the following packages.

library(tidyverse)
library(multcomp)

Running the simple lm model

For this example I will use the classic data from the cars library

#loading the source code
source("Improving_MultComp/Categorize_lm.R")

# Example usage
lm_model <- lm(mpg ~ wt+hp+qsec+vs+am, data = mtcars)
summary(lm_model)
View(categorize_association(lm_model))

lm_model <- lm(mpg ~ -1 +wt+hp+qsec+vs+am, data = mtcars)
summary(lm_model)
View(categorize_association(lm_model))

# Output: Detailed summary with classifications for each coefficient

As you can see the output now extracted the p-value and the confidence interval for the estimates of every regression coefficient. Then, we can look at its range and create a classification for each estimate from the linear model.

We can also make a plot out of this output

source("Improving_MultComp/plot_association.R")
plot_association(lm_model)

Next step: made the code to also include mixed models

Multiple comparisons

As I said I also want to make it so we can extract the same type of classification from multiple comparisons. For that I have choosen to start with the multcomp package

source("Improving_MultComp/Categories_glht.R")

### multiple linear model, swiss data
lmod <- lm(Fertility ~ ., data = swiss)

### test of H_0: all regression coefficients are zero 
### (ignore intercept)

### define coefficients of linear function directly
K <- diag(length(coef(lmod)))[-1,]
rownames(K) <- names(coef(lmod))[-1]
K

### set up general linear hypothesis
lmod.glht <- glht(lmod, linfct = K)
str(lmod.glht)
summary(lmod.glht)
View(extract_glht_ci(lmod.glht))
plot_glht_ci(lmod.glht)
### alternatively, use a symbolic description 
### instead of a matrix 
lmod.glht2 <- glht(lmod, linfct = c("Agriculture = 0",
                      "Examination = 0",
                      "Education = 0",
                      "Catholic = 0",
                      "Infant.Mortality = 0"))
extract_glht_ci(lmod.glht2)
plot_glht_ci(lmod.glht2)

### multiple comparison procedures
### set up a one-way ANOVA
amod <- aov(breaks ~ tension, data = warpbreaks)

### set up all-pair comparisons for factor `tension'
### using a symbolic description (`type' argument 
### to `contrMat()')
lmod.glht3 <- glht(amod, linfct = mcp(tension = "Tukey"))
extract_glht_ci(lmod.glht3)
plot_glht_ci(lmod.glht3)

### alternatively, describe differences symbolically
lmod.glht4 <- glht(amod, linfct = mcp(tension = c("M - L = 0", 
                                    "H - L = 0",
                                    "H - M = 0")))
extract_glht_ci(lmod.glht3)
plot_glht_ci(lmod.glht3)

### mix of one- and two-sided alternatives
warpbreaks.aov <- aov(breaks ~ wool + tension,
                      data = warpbreaks)

### contrasts for `tension'
K <- rbind("L - M" = c( 1, -1,  0),     
           "M - L" = c(-1,  1,  0),       
           "L - H" = c( 1,  0, -1),     
           "M - H" = c( 0,  1, -1))

warpbreaks.mc <- glht(warpbreaks.aov, 
                      linfct = mcp(tension = K),
                      alternative = "less")
confint(warpbreaks.mc, level = 0.95)$confint
extract_glht_ci(warpbreaks.mc)
plot_glht_ci(warpbreaks.mc)