In jcastagna/ggDiagnostic: Diagnostic Plots in ggplot2

knitr::opts_chunk$set(echo = TRUE, message=FALSE,warning=FALSE)

Why?

Save some time. Quick & Dirty model diagnostic plots in ggplot.

Plan

1. Redo base plots in ggplot

par(mfrow=c(2,3))
lm.1 <- lm(mpg ~ wt,data=mtcars)
plot(lm.1,which=1:6)

2. Create ggplot versions ~ make tidy

library(tidyverse)
library(broom)

My.Mod <- lm.1

#Base
plot(My.Mod,which=1)

#GGplot
D1  <- augment(My.Mod) %>% ggplot(aes(x=.fitted,y=.std.resid)) +
  geom_point() + 
  geom_smooth(se=FALSE,colour="red",size=.25) +
  geom_hline(yintercept=0,linetype=3) +
  labs(title="Residuls vs Fitted",subtitle=My.Mod$call)

D1

3. Functionalise

How to determine models?
What sort of arguments, functionality?
PAss through
Design ?

## What should it look like?

# Pehaps???

MyModel %>% ggdiag()

Data %>% MyModel %>% ggdiag()

# plot index no or rather name?
Data %>% MyModel %>% ggdiag(Plot=1:2)
Data %>% MyModel %>% ggdiag(Plot=c("RVFit","qq"))

4. Ensure plays nice with existing

x. Other Thoughts Practice with RProj, Git.
Don't re-invent the wheel - just make quicker to change the flat.

Questions ???

Logic for text annotation? How determine cutoff. Starting assumptiom? Start from a) model (eg lm) or b) tidy output of model
What are the statistics that determine highlighting text labels in base plots?

# How do you see how current plot is coded?
# This doesnt help
plot

jcastagna/ggDiagnostic documentation built on May 30, 2019, 4:34 p.m.