Changing size of legend and point shapes in plot from CGPfunctions::Plot2WayANOVA with GGplot2 in R

The problem here is that Plot2WayANOVA is a wrapper function around ggplot that attempts to make plotting your ANOVA easier by handling many of the technical aspects of creating the appropriate ggplot to visualize the data. Although this kind of function is good for learners or for quick visualisation of data, it doesn't lend itself to customizing a plot the way you want.

The option to do all the things you ask doesn't exist within the function itself. You can pass extra geom calls and scale calls as well as theme calls, but the difficulty is that the shape aesthetic for the points is not mapped. It is possible to add a theme that widens the legend key, but when you do so you find that the line for the errobars is also included in the legend as a thin unbroken line overlying the thick patterned lines, which looks messy and again cannot be changed through the function parameters.

You therefore have two options:

1. Build the plot yourself using ggplot directly : the problem with this is that it would be a pain to get all of the relevant model summary information into the caption below the title, and of course defeats the point of having Plot2WayANOVA in the first place
2. Modify the plot object returned by Plot2WayANOVA : The problem with this is that Plot2WayANOVA doesn't actually return a ggplot object; it only prints the ggplot, then returns a nice big list of information about the analysis. You will need to hack the function to get it to return a ggplot object.

I will demonstrate how you would go about the second method here. It requires a bit less code than building from scratch, but is also a bit more technical and harder to follow.

First, write a new version of Plot2WayANOVA that returns a ggplot instead of the summary object:

f <- CGPfunctions::Plot2WayANOVA
body(f)[[84]] <- quote(return(p))
My_Plot2WayANOVA <- f

So now we can get our ggplot object by doing:

p <- My_Plot2WayANOVA(formula = mpg ~ am * cyl, dataframe = mtcars)

We will need to hack at the aesthetic mappings to get them to behave as we want, and overwrite the geom_errorbar to ensure it has a blank key glyph:

p$layers[[3]]$aes_params$shape   <- NULL
p$layers[[3]]$aes_params$colour  <- NULL
p$layers[[3]]$geom_params$colour <- NULL
p$layers[[3]]$geom_params$shape  <- NULL
p$layers[[3]]$mapping            <- aes(color = cyl, fill = cyl)
p$mapping                        <- aes(colour = cyl, shape = cyl, 
                                        fill = cyl, group = cyl,
                                        x = am, y = TheMean)

p$layers[[1]] <- geom_errorbar(aes(ymin = LowerBound, ymax = UpperBound), 
                               size = 1, width = 0.1, key_glyph = "blank")

Now p is a ggplot object to which you can add custom scales and themes as you see fit:

p + scale_shape_manual(values = c(21:23), name = "cyl") +
    theme(legend.key.width = unit(75, "points"))

Whether it is worth going to all this trouble is debatable; it would probably be easier and safer to build the plot yourself from scratch if you really find the original styling unacceptable.

EDIT

Here's a working example with the data added to the question:

p <- My_Plot2WayANOVA(formula = CigAd ~ Wave * storetype, dataframe = d)

p$layers[[3]]$aes_params$shape   <- NULL
p$layers[[3]]$aes_params$colour  <- NULL
p$layers[[3]]$geom_params$colour <- NULL
p$layers[[3]]$geom_params$shape  <- NULL
p$layers[[3]]$mapping <- aes(color = storetype, fill = storetype)
p$mapping <- aes(colour = storetype, shape = storetype, 
                 fill = storetype, group = storetype,
                 x = Wave, y = TheMean)
p$layers[[1]] <- geom_errorbar(aes(ymin = LowerBound, ymax = UpperBound), 
                               size = 1, width = 0.1, key_glyph = "blank")

p + scale_shape_manual(values = c(21:23), name = "storetype") +
    theme(legend.key.width = unit(75, "points"))