Creating a "normalized share over time" plot in R

Question

I'm interested in making a plot that is normalized and can trend share over time. Here is an example:

(http://stevecoast.com/wp-content/uploads/2012/02/normalised-phone-share2-001.jpg)

The data I am using is just 2 factors (1,0), so there would be two colors. There are 3178 observations in total. I'm not sure if there is a function that will allow me to keep the data in this form or transformation will be required.

set.seed(124)
variableValue <- sample(0:1, 20, replace = TRUE)
set.seed(124)
timePeriod <-sort(sample(letters[1:5], 20, replace=TRUE))
sort(timePeriod)
data<-data.frame(variableValue,timePeriod)
data

I figured ggplot is the best way to go, but I'm pretty lost as to where to start.

Any advice would be awesome. Thanks.

Start with sharing (or simulating) data to make a [reproducible example](http://stackoverflow.com/q/5963269/903061). — Gregor Thomas, Oct 08 '15 at 23:49
Do you really have a categorical variable for your x-axis? Not a date or time? — Gregor Thomas, Oct 09 '15 at 00:52

Gregor Thomas · Accepted Answer · 2015-10-09T16:22:16.403

Since you want variableValue treated categorically, we'll first convert it to a factor:

data$variableValue = factor(data$variableValue)

You can do a lot of data manipulation inside ggplot, but I prefer to do it beforehand for better transparency.

library(dplyr)
dat_summ = data %>% group_by(timePeriod) %>%
    mutate(n_time = n()) %>%
    group_by(timePeriod, variableValue) %>%
    summarize(proportion = n() / first(n_time))

This makes a data frame with one row per variableValue per timePeriod, and a proportion column for each: exactly what we want to plot.

ggplot(dat_summ, aes(x = timePeriod, y = proportion,
                     fill = variableValue, group = variableValue)) +
    geom_area() +
    scale_y_continuous(labels = scales::percent)

Then we can plot. We specify the variables that get mapped to x and y axes and fill color, and since we have a categorical variable on the x-axis we need to specify a group definition to "connect the dots". geom_area is a filled-in area graph, it's default will stack the areas on top of each other like we want. To be fancy, I specify to use percent scales on the y-axis---this whole line could be omitted otherwise.

After copying and pasting your code - I keep getting an error: Variable 'n_total' not found.....Am I missing a package? — Mark Romano, Oct 09 '15 at 15:47
@MarkRomano no, sorry. I renamed `n_total` to `n_time` but didn't make the change both places it showed up. Should work now. — Gregor Thomas, Oct 09 '15 at 16:22
Gotcha....I'm not too familiar with dplyr package (though I should!) and thought maybe it was a special keyword. Worked like a charm, thanks! — Mark Romano, Oct 09 '15 at 17:17

Creating a "normalized share over time" plot in R

1 Answers1