2) Fonts. Picking fonts can be really tricky, but there are some really great resources out there (see below, where we're back to our "let others help you" mantra!).
Here, I've simply applied the fonts from my own website, changing the family element of element_text().
3) Text size. You can manipulate text size within theme() either by setting absolute sizes (e.g. size = 16), or relative sizes (e.g. size = rel(1.2)).
The relative size is a good idea if you're going to reuse this theme: change the base size as needed and everything follows!
At this point, we've done most of the work, but we can still make our data story easier to take in by giving everything a bit more space to breathe.
First, let's move the legend to reduce unnecessary eye movements, fade the grid, and remove an unnecessary axis title.
And now let's change the base size, increase the default line height and add margins around the text elements and around the plot itself, using the acronym TRouBLe to remember the margin values go Top, Right, Bottom, Left (or clockwise from the top).
Tada!
Finally, let's package this up as a function we can easily reuse to create consistency across our project by just adding `+ theme_rladiesdemo()` to every plot.
Even with no effort to reconcile colours (see yesterday's thread for tips on that), it has a big impact!
And now for some resources (1/3):
- Font advice: @glyphe. Oliver runs a great weekly newsletter with font recommendations, and has some really practical resources on his site, including a free webinar from today which should be available to replay later. pimpmytype.com/font-choice/
Day 3: Writing functions to create parameterised graphs
So, we've picked our colours and set up a nice theme. We already know we want to apply this to all our plots but what if we could reuse the same plot code across different bits of the data?
Easy, let's make a function!
First, let's set up our plot. We're going to plot the number of penguins from each species within our dataset. We're using ggchicklet::geom_chicklet(), an anchor colour which is a blend of the blue and purple I mentioned on Day 1 and theme_rladiesdemo() which we built on Day 2.
Now let's add some labels, so we can state the number of penguins in each species, along with the mean body mass.
To do this, I'm using my go-to #rstats annotation package {ggtext}, which allows us to apply some CSS to create the text hierarchy principles from Day 2.
Here are three reasons why I think you should do this:
- Help orient your readers with text hierarchy
- Give everything some space to breathe
- Achieve effortless consistency with one extra line of code
Sound good? Let's dig in!
My starting point for creating a custom theme is typically theme_minimal(). It has sensible defaults such as relative text size and margins that we can build on, by just replacing some elements.
🤫 I'm going to let you in on a secret... I find picking colours really tricky! Thankfully, I've found few ways round that.
My top tip is to let others help you! But first, a broad principle...
When picking colours for story telling, I try to make the colours as intuitive as possible.
Here's the adventure I took the Palmer Penguins on in a recent talk involving the #GreatPenguinBakeOff. See if you can guess the details. (The next tweet should give you a few clues!)
It's not about making your plots into a guessing game. It's about reducing cognitive load by making it easy to remember what's what.
And this allows me to illustrate one way to let others help you: photos! All of the colours in the previous plots were taken from these photos.
Hi folks! I'm Cara, an Edinburgh-based freelance data consultant, and I'm excited to be bringing you a week of #rstats content based around the things I enjoy creating the most: data visualisations and "enhanced" reproducible outputs.
Over the course of the week, we'll be exploring the different components that go into making a nice "branded" parameterised report, each day building on the days before until we get to our finished product: a parameterised celebration of RLadies within the NHS.
Here's the menu:
- Day 1: Setting up a colour scheme
- Day 2: Building a custom ggplot theme
- Day 3: Writing functions to create parameterised graphs
- Day 4: Manipulating text with R
- Day 5: A worked example of designing a pdf template (and a case for PDFs)
- Day 6: Our finished product!
Today I'd like to share how I debug R code, highlighting some neat RStudio tools and tricks. It's not a linear process so this thread isn't exactly in order. 1/n
It may sounds silly, but first, I take a breath & *fully* read the error/warning message. The breath is because it may not be the first error of the day & I might be getting frustrated. The careful reading is because I want all available information before moving ahead. 2/n
I think about if I've seen the error before. My most common errors are easily fixed & include
object 'X' not found -> I forgot to load X or misspelled it
unexpected ')' -> I have an extra ) somewhere. Also for ] and }
3/n
This morning, I'll share some tips for my favorite thing to do in R: make plots! I'm firming in the ggplot world so many tips are related to this and related packages.
So not surprisingly, my first tip is to checkout ggplot! I think the most powerful aspect it it's legends as they are created automatically and change as you change the plot. No fear of an incorrect legend!
ggplot is in the tidyverse & thus, works seamlessly with dplyr, tidyr, etc. If I need to transform data *just* for a plot, I use pipes (%>%) to avoid saving the data and cluttering up my environment. For example: