How to View a Summary of a Data Frame in Julia
Guide1
When handling data frames in Julia, you can use the describe function to quickly check basic statistics. This plays a role similar to the summary function in R. As an example, suppose the following data frame is given.
julia> data = DataFrame(x = rand(10), y = 1 .+ randn(10), z = 10randn(10))
10×3 DataFrame
Row │ x y z
│ Float64 Float64 Float64
─────┼───────────────────────────────
1 │ 0.581684 0.613567 -6.96196
2 │ 0.167714 1.86029 2.25717
3 │ 0.6753 0.397726 -6.88198
4 │ 0.605366 1.27918 16.082
5 │ 0.398977 0.655179 -3.7648
6 │ 0.283162 1.40468 15.9616
7 │ 0.366776 0.680256 11.8967
8 │ 0.387233 2.15353 -3.79308
9 │ 0.556446 1.58978 -11.0116
10 │ 0.344502 0.894148 18.8104
Here, $X \sim U (0, 1)$, $Y \sim N(1, 1)$, and $Z \sim N(0, 10)$.
Basic Statistics
julia> describe(data)
3×7 DataFrame
Row │ variable mean min median max nmissing eltype
│ Symbol Float64 Float64 Float64 Float64 Int64 DataType
─────┼─────────────────────────────────────────────────────────────────────────
1 │ x 0.436716 0.167714 0.393105 0.6753 0 Float64
2 │ y 1.15283 0.397726 1.08667 2.15353 0 Float64
3 │ z 3.25945 -11.0116 -0.753812 18.8104 0 Float64
Using the describe function without any options lets you check, for each column, the mean, minimum, median, maximum, number of missing values, and data type, as shown above.
Specific Statistics
julia> describe(data, :mean, :std)
3×3 DataFrame
Row │ variable mean std
│ Symbol Float64 Float64
─────┼───────────────────────────────
1 │ x 0.436716 0.161219
2 │ y 1.15283 0.593586
3 │ z 3.25945 11.3199
You can also view only the statistics you want. The candidates provided by default are :mean, :std, :min, :q25, :median, :q75, :max, :sum, :eltype, :nunique, :nuniqueall, :first, :last, :nnonmissing, and :nmissing, and using :all lets you check every statistic listed.
User-Defined Statistics
julia> describe(data, rand => :rand)
3×2 DataFrame
Row │ variable rand
│ Symbol Float64
─────┼─────────────────────
1 │ x 0.556446
2 │ y 0.680256
3 │ z -3.79308
It does not necessarily have to be a statistic; it seems that any function that takes a vector and returns a scalar can be used. If you give a pair in the form function => :name, the function function is applied to every column and the result remains under the column name :name. In the example above, one value was randomly drawn from each column.
Complete Code
using DataFrames
data = DataFrame(x = rand(10), y = 1 .+ randn(10), z = 10randn(10))
describe(data)
describe(data, :mean, :std)
describe(data, rand => :rand)
