More standard ggplot2 behavior

[antaldaniel]

Excellent work, I have been thinking for some time to do the same. I think that this is very good groundwork for a widely used geometry, but I find it a bit too specific. I have two issues.

a) I think that any not necessary adjustment, including theming, coloring, coord_flip() atc should be left exactly like in the case of geom_boxplot () to the user.

For demo purposes I think that the Economist theme is great, because that is what everybody knows and envies, but people will use the geom widely if it fits smoothly into their existing reports, documents, charts. Flipping the coords is important for demonstrating the Economist origin, but it is a standard function and can be confusing because it changes the x, y values.

b) for me a more logical usage would be

ggplot ( data = dat,
aes ( x = numeric_var, minpoint = ( quantile (numeric_var, .1, na.rm=TRUE), midpoint = ( median(numeric_var, na.rm=TRUE), maxpoint = ( quantile (numeric_var, .9, na.rm= TRUE), group = metro_area)) +
geom_econdist( aes ( mincolor = 'lightblue', midcolor = 'grey10', maxcolor = 'blue') +
coord_flip() +
labs () +
theme ()

It would be a more flexible version of the geom_boxplot (), where the quantile values cannot be moved, and the midpoint can take mean, average, or some weighted average value. This would allow both already computed tables to be used, as in your example ,or a geom_boxplot() like charting.