milesmcbain / visdat

A package to assist in visually testing a dataset

visdat

What does visdat do?

Initially inspired by csv-fingerprint, vis_dat helps you visualise a dataframe and "get a look at the data" by displaying the variable classes in a dataframe as a plot with vis_dat, and getting a brief look into missing data patterns vis_miss.

The name visdat was chosen as I think in the future it could be integrated with testdat. The idea being that first you visualise your data (visdat), then you run tests from testdat to fix them.

There are currently two main commands: vis_dat and vis_miss.

• vis_dat visualises a dataframe showing you what the classes of the columns are, and also displaying the missing data.

• vis_miss visualises just the missing data, and allows for missingness to be clustered and columns rearranged. vis_miss is similar to missing.pattern.plot from the mi package. Unfortunately missing.pattern.plot is no longer in the mi package (well, as of 14/02/2016).

• new!! vis_guess has a guess at what the value of each cell. So "10.1" will return "double", and 10.1 will return "double", and 01/01/01 will return "date". Keep in mind that it is a guess at what each cell is, so you can't trust this fully. vis_guess is made possible thanks to Hadley Wickham's readr package - thanks mate!

How to install

# install.packages("devtools")

library(devtools)

install_github("tierneyn/visdat")

Example

Let's see what's inside the dataset airquality

library(visdat)

vis_dat(airquality)

The classes are represented on the legend, and missing data represented by grey.

by default, vis_dat sorts the columns according to the type of the data in the vectors. You can turn this off by setting sort_type == FALSE. This feature is better illustrated using the example2 dataset, borrowed from csv-fingerprint.

vis_dat(example2)

vis_dat(example2, 
        sort_type = FALSE)

The plot above tells us that R reads this dataset as having numeric and integer values, along with some missing data in Ozone and Solar.R.

We can explore the missing data further using vis_miss

vis_miss(airquality)

You can cluster the missingness by setting cluster = TRUE

vis_miss(airquality, 
         cluster = TRUE)

The columns can also just be arranged by columns with most missingness, by setting sort_miss = TRUE.

vis_miss(airquality,
         sort_miss = TRUE)

vis_guess takes a guess at what each cell is. It's best illustrated using some messy data, which we'll make here.

messy_vector <- c(TRUE,
                  T,
                  "TRUE",
                  "T",
                  "01/01/01",
                  "01/01/2001",
                  NA,
                  NaN,
                  "NA",
                  "Na",
                  "na",
                  "10",
                  10,
                  "10.1",
                  10.1,
                  "abc",
                  "$%TG")

messy_df <- data.frame(var1 = messy_vector,
                       var2 = sample(messy_vector),
                       var3 = sample(messy_vector))

vis_guess(messy_df)

So here we see that there are many different kinds of data in your dataframe. As an analyst this might be a depressing finding. Compare this to vis_dat.

vis_dat(messy_df)

Where you'd just assume your data is wierd because it's all factors - or worse, not notice that this is a problem.

At the moment vis_guess is slow as - take this into consideration when you are using it on data with more than 1000 rows. We're looking into ways of making it faster, potentially using methods from the parallel package, or writing some c++ code.

Future work

visualising expectations

Another idea is to pass expectations into vis_dat or vis_miss, along the lines of the expectation command in assertr. For example, you could ask vis_dat to identify those cells with values of -1 with something like this:

data %>% 
  expect(value == -1) %>%
  vis_dat

vis_datly. vis_dat could include an interactive version of the plots using plotly, so that you can actually see what is inside the data.

Thank you

Thank you to Jenny Bryan, whose tweet got me thinking about vis_dat, and for her code contributions that remove a lot of testing errors.

Thank you to Hadley Wickham for suggesting the use of the internals of readr to make vis_guess work.

Thank you to Miles McBain for his suggestions on how to improve vis guess. This resulted in making it at least 2-3 times faster.

Thanks also to Noam Ross for his suggestions and code for using plotly with visdat.