regresion_diagnostics.R

# Model Diagnostics

**Reference only, remove from report before sending to the sponsor**

### Influential Observations

```{r influence_plot, out.extra = "figure", echo = F, warning = F, message = F, comment = NA, eval = F}
influencePlot(mods$iris_data, id.method = "identify", main = "Influence Plot", sub = "Circle size is proportial to Cook's Distance" )

```

### Linearity of Relationships

```{r added_variable_plot, out.extra = "figure", echo = F, warning = F, message = F, eval = F}
avPlots(mods$iris_data)

```

```{r nonlinearity_crPlots, out.extra = "figure", echo = F, warning = F, message = F, eval = F}
crPlots(mods$iris_data) # component + residual plot 

```

### Consitancy of Error Variance (homoscedasticity)

```{r homoscedasticity_test, echo = F, comment = NA, eval = F}
ncvTest(mods$iris_data) # non-constant error variance test

```

```{r homoscedasticity_plot, out.extra = "figure", echo = F, warning = F, message = F, comment = NA, eval = F}
# plot studentized residuals vs. fitted values 
spreadLevelPlot(mods$iris_data)

```

### Normality of Errors

```{r qqplot, out.extra = "figure", echo = F, warning = F, message = F, comment = NA, eval = F}
qqPlot(mods$iris_data, main="QQ Plot")

```

```{r residuals_distribution, out.extra = "figure", echo = F, warning = F, message = F, eval = F}
sresid <- MASS::studres(mods$iris_data) 
hist(sresid, freq=FALSE, 
   main="Distribution of Studentized Residuals")
xfit<-seq(min(sresid),max(sresid),length=40) 
yfit<-dnorm(xfit) 
lines(xfit, yfit)

```

### Collinearity

```{r Collinearity, table = T, results = "hide", eval = F}
vif(mods$iris_data) %>% as_tibble() # variance inflation factors 

```

$\sqrt(model) > 2$

```{r Collinearity_sq, table = T, results = "hide", eval = F}
vif(mods$iris_data) %>% as_tibble() %>% mutate_all(list(~sqrt(.) > 2)) # problem?

```