Faceting in ggplot() in R

Question

I am trying to build a plot for a numeric variable rider_count vs a categorical variable weekdays("Mon", "Tue"....), and this plot is required to be a faceting plot with 55 categories,

I tried to use

ggplot(aes(x=wday, y=rider_count_sum)) +
  geom_bar(stat = "identity") +
  facet_wrap(~counter_edited, scales="free")

However, the output of it is twisted very hard due to the scale does not fit.

Are there any ways to make it scale normally?

Answer 1

The issue you here is your faceting. It produces a grid of 8 x 7 cells. The plot displays on my monitor at about 18cm x 11cm in size. That means each cell is approximately 2.25cm x 1.5cm. Is a cell of that size large enough to provide meaningful information in the form of a plot? I would say "no".

So, you have two options: increase the size of the graphic or reduce the size of the grid.

Is increasing the size of the plot an option? Well, how big would each cell have to be to be meaningful? I don't know: you'd have to experiment: it would depend on the viewing distance and the level of information you'd want to convey. As a thought experiment, let's say you need each cell to be 8cm x 8cm to be interpretable. That means the graphic would need to be at least 64cm x 56cm. That would require an A1/ANSI D sheet of paper. That's heading to paper size. Unless you're talking posters, that's not reasonable. Even as a poster, a reader would have to stand so close that they wouldn't get the message of the whole graphic. On a digital display, you'd again be talking about a wall mounted unit. Standing close enough to look at a cell, pixel resulution would be an issue. Scrolling on a smaller unit would destroy the whole purpose of using a facted display.

Pagination would also destroy the benefit of faceting: you wouldn't be able to see all the data at the same time.

So, whilst increasing the size of your plot might be technical possible, I don't think it would be practically useful.

What about reducing the number of cells? That to me would be the way to go. Simplify your presentation to allow your message to come across. For example, summairse weekdays vs weekends in one graphic, differences between weekdays in another. That reduces one dimension from 7 to either 2 or 5. I don't know how you construct counter_edited, so I don't know what the columns of your facet represent, but could you perhaps reduce the number of categories to 3 or 4? Combined with my weekday/weekend suggestion, would give you grids of between 4x5 and 2x3. Much more managable (though even 4x5 may be too complex).

In short: even if making you current graphic look better is technically possible, I doubt it will ever be practically useful. I suggest adopting a different approach. The question I would ask is deeper than the simple technical one of improving your graphic: what is your underlying purpose? Once you know that, adapt your presentation to best address your objective.