DataFramesMeta selecting multiple columns

New to Julia
1

Using DataFramesMeta (or others), can I select multiple columns of choice and do some calculations?

Here’s a mock data. I would like to subset by column x, and then calculate column-wise means. Here’s the answer.

df = DataFrame(x = [1,1,1,2,3,2,3,2,3,], y1 = [2,1,2,1,2,1,2,1,2], y2 = [5,7,6,5,7,6,5,7,6], y3 = [4,2,3,4,2,3,4,2,3])
l = [:y1, :y2, :y3]
by(df, :x, d->mean(Array{Float64,2}(d[l]); dims=1))

3×4 DataFrame
│ Row │ x │ x1 │ x2 │ x3 │
├──┼─┼────┼───┼──┤
│ 1 │ 1 │ 1.66667 │ 6.0 │ 3.0 │
│ 2 │ 2 │ 1.0 │ 6.0 │ 3.0 │
│ 3 │ 3 │ 2.0 │ 6.0 │ 3.0 │

I also wanted to do this using @linq so that it can be coupled with other functions. But, mine didn’t work (obviously).

@linq df |> by(:x, mean(Array{Float64,2}(l; dims=2)))

I wanted to,

  1. subset a data frame by column x
  2. select columns of choice (y1, y2, y3)
  3. calculate column-wise means for y1, y2, y3

Any idea?

Thanks,

2

Can you clarify what you mean by “select multiple columns of choice and do some calcuations”?

It’s unclear why you have Array{Int,2} in your function call, as well. What are you trying to achieve?

3

@pdeffebach sorry about that. I edited my post. I think here’s a better representation.

I wanted to

  1. subset a data frame by column x
  2. select columns of choice (y1, y2, y3)
  3. calculate column-wise means for y1, y2, y3
4

Ah I see. What you are asking for is difficult to do with DataFramesMeta. Better to use Base DataFrames instead:

cols = [:x2, :x3, :x4]
by(df, :a) do d
       DataFrame([mean(d[col]) for col in cols]', cols)
end
5

@pdeffebach Thanks!

6

colwise and aggregate are also useful for this operation