In riinuots/tidyforest: A tidyverse workflow for running and presenting a logistic regression model in R.

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-"
)

tidyforest

tidyforest contains a tidyverse workflow for running and presenting a logistic regression model in R. This is organised into 5 functions:

• melanoma_formatting() - renaming/creating factor levels. This formats the boot::melanoma dataset which can then be used as example input for the following functions.
• glm_tidyresult() - running a logistic regression model and formatting the output into a nice dataframe (glm - generalized linear models).
• or_plot() - creates a forest plot (or - odds ratios).
• or_table_plot() - creates a graphical table aligned exactly to the forest plot.
• tidyforest() - combines all of the above, and puts or_plot() and or_table_plot side by side.

Install

install.packages("devtools")

and

devtools::install_github("riinuots/tidyforest")

Use

library(tidyforest)

mydata = data_melanoma()
dependent = 'status.factor' 
explanatory = c('sex.factor', 'ulcer.factor', 'stage.factor')
explanatory_names = c('Gender', 'Ulceration', 'Stage (thickness)')

tidyforest(mydata, dependent, explanatory, explanatory_names)

Note that outcome variable ("dependent") actually includes 3 levels: "Alive", "Died", and "Died of other causes". This example lumps the latter categories into once which might not be desired. You should always plot your data in an exploratory way (e.g. boxplots, barplots) before running advanced statistical models on it.

These warnings:

Warning messages:
1: Removed 21 rows containing missing values (geom_text). 
2: Removed 3 rows containing missing values (geom_errorbar). 
3: Removed 3 rows containing missing values (geom_point). 
4: Removed 3 rows containing missing values (geom_text). 

can safely be ignored, as they refer to the fact that the reference levels do not have a point and an errorbar (does not make sense), or that the explanatory_names are only printed once for the whole group (not on each line).

Or call the functions separetely

library(tidyforest)

mydata = data_melanoma()

mydata %>% str() #prints the structure and first few observations

dependent = 'status.factor'
explanatory = c('sex.factor', 'ulcer.factor', 'stage.factor')
explanatory_names = c('Gender', 'Ulceration', 'Stage (thickness)') #this is optional

glm_result = glm_tidyresult(mydata, dependent, explanatory)

or_plot(glm_result)

or_table_plot(glm_result, explanatory_names)

riinuots/tidyforest documentation built on May 27, 2019, 8:51 a.m.