IrisDataLinearModel.Rmd

IRIS DATA SET LINEAR MODELLING
========================================================


```{r}
data(iris)
linearModel = lm(iris$Sepal.Length ~ iris$Sepal.Width)
unexplainedByModel = linearModel$residuals
```

Now we can make some plots of the other variables and how they relate to the stuff not explained by the linear model
```{r}
plot(iris$Petal.Length,unexplainedByModel)

```

#note gap and a trend for greater petal length to explain what was not explained by the lm
```{r}
plot(iris$Petal.Width,unexplainedByModel)
```

#note gap and a trend for greater petal length to explain what was not explained by the lm, but less tight than Petal.Length
```{r}

boxplot(unexplainedByModel ~ iris$Species)
```

#those are some pretty significant differences as all the central boxes are clear of each other

So at the moment, it looks like Species is really, really important. If we want to see the differences Petal.Length, but colour the results by species, we could do that. There are a couple of ways, but I like to build my graphs up from basic principles over a bunch of steps:

#we are going to need to get the axis scales formally set for doing a multiple plot
```{r}

maxy= max(unexplainedByModel)
miny= min(unexplainedByModel)
maxx= max(iris$Petal.Length)
minx= min(iris$Petal.Length)

```

#have enough colours for everything we are going to plot
```{r}

colist = c("#FF000033","#00FF0033","#0000FF33")

```

#our base graph
```{r}

rowcriteria = iris$Species=="setosa"
plot(iris$Petal.Length[rowcriteria], jitter(unexplainedByModel[rowcriteria], factor=0.3), col= colist[1], pch=19, xlim=c(minx,maxx), ylim=c(miny,maxy), frame.plot=FALSE, xlab="Petal Length", ylab="Residuals")
```
#we add our next set of points
```{r}
rowcriteria = iris$Species=="versicolor"
points(iris$Petal.Length[rowcriteria], jitter(unexplainedByModel[rowcriteria], factor=0.3), col= colist[2], pch=19)

```

#and we add our final set
```{r}
rowcriteria = iris$Species=="virginica"
points(iris$Petal.Length[rowcriteria], jitter(unexplainedByModel[rowcriteria], factor=0.3), col= colist[3], pch=19)

```


#now let's add a legend
```{r}
labels <- c('setosa', 'versicolor', "virginica")
position <- 'topleft'
inset <- c(0.02, 0)
legend(position, labels, fill=colist, inset=inset)
```