グループ化集計統計

Julia でのデータ操作

Katerina Zahradova

Instructor

ここまでの要点

# 最低賃金の平均を計算
combine(wages, :effective_min_wage_2020_dollars => mean)

1x1 DataFrame
Row | effective_min_wage_2020_dollars_mean
    | Float64
____|_______________________
   1| 8.37093

ここまでの要点

# フィルタして平均を計算
first(combine(filter(r -> r.region =="W", wages), :effective_min_wage_2020_dollars => mean))
first(combine(filter(r -> r.region =="S", wages), :effective_min_wage_2020_dollars => mean))
first(combine(filter(r -> r.region =="NE", wages), :effective_min_wage_2020_dollars => mean))

DataFrameRow (1 columns)
Row | effective_min_wage_2020_dollars_mean
    | Float64
____|_______________________
   1| 8.75413
...

combine() と groupby() の使用

# 地域でグループ化
wages_by_region = groupby(wages, :region)

# グループごとに平均を計算
combine(wages_by_region, :effective_min_wage_2020_dollars => mean)

4x2 DataFrame
Row region    effective_min_wage_2020_dollars_mean
    String    Float64
__________________________
1    S        8.15458
2    W        8.59119
3    NE       8.75413
4    MW       8.1514

combine() と groupby() を組み合わせる

# まとめて実行
combine(groupby(wages, :region), :effective_min_wage_2020_dollars => mean)

4x2 DataFrame
Row region    effective_min_wage_2020_dollars_mean
    String    Float64
__________________________
1    S        8.15458
2    W        8.59119
3    NE       8.75413
4    MW       8.1514

combine() と groupby() の使用

# 列名を変更
combine(groupby(wages, :region), 
        :effective_min_wage_2020_dollars => mean => :average_min_wage_2020_dollars)

4x2 DataFrame
Row region    average_min_wage_2020_dollars
    String    Float64
__________________________
1    S        8.15458
2    W        8.59119
3    NE       8.75413
4    MW       8.1514

1列に複数の関数

# 1列に複数の関数を適用
combine(groupby(wages, :region), 
                    :effective_min_wage_2020_dollars .=> [mean, median, maximum])

4x4 DataFrame
Row  region  effective_min_wage_2020_dollars_mean  ...
     String  Float64                               ...
____________________________________________________________________________________
1    S       8.15458                               ...
2    W       8.59119                               ...
3    NE      8.75413                               ...
4    MW      8.1514                                ...

1列に複数の関数

# 1列に複数の関数を適用
combine(groupby(wages, :region), :effective_min_wage_2020_dollars .=> [mean, median] .=> [:average, :median])

4x2 DataFrame
Row region  average  median
    String  Float64  Float64
____________________________
1    S      8.15458  8.0
2    W      8.59119  8.34
...

# ドットを忘れない！
combine(groupby(wages, :region), :effective_min_wage_2020_dollars => [mean, median])

ArgumentError: Unrecognized column selector ...

複数列に1つの関数

combine(groupby(wages, :region), [:state_min_wage, :federal_min_wage] .=> mean)

4x2 DataFrame
Row region  state_min_wage_mean  federal_min_wage_mean
    String  Float64              Float64
______________________________________________________
1   S       2.73128              4.35566
2    W       4.26638              4.35566
...

# ドットを忘れない
combine(groupby(wages, :region), [:state_min_wage, :federal_min_wage] => mean)

MethodError: objects of type ...

複数列に複数の関数

# 関数を1行行列で指定
combine(groupby(wages, :region), [:state_min, :federal_min] .=> [mean median])

Row region  state_min_mean  federal_min_mean  state_min_median  federal_min_median
_________________________________________________________________________________
1   S       2.73128         4.35566           2.0                4.25
...

# 関数をベクタで指定
combine(groupby(wages, :region), [:state_min, :federal_min] .=> [mean, median])

Row region  state_min_mean  federal_min_median 
______________________________________________
1   S       2.73128         4.25
...

利用可能な関数

使用可能な関数:

sum(), mean(), minimum() などの基本統計
ユーザー定義関数（ブロードキャスト）
無名関数（ByRow()で包む）
DataFrames の特別な関数: nrow, proprow など

チートシート

# Grouped DataFrame gdf
# 1列 + 1関数
combine(gdf, :c => f => :new_c)

# 1列 + 2つ以上の関数
combine(gdf, :c .=> [f1, f2, ...] .=> [:new_c_f1, :new_c_f2, ...])

# 2列以上 + 1関数
combine(gdf, [:c1, :c2, ...] .=> f .=> [:new_c1_f, :new_c2_f, ...])

# 2列以上 + 2つ以上の関数 - 全組合せ
combine(gdf, [:c1, :c2, ...] .=> [f1 f2 ...] .=> [:c1_f1, :c2_f1, ..., :c1_f2, ...])

# 2列以上 + 2つ以上の関数 - 対応づけ
combine(gdf, [:c1, :c2, ...] .=> [f1, f2, ...] .=> [:new_c1_f1, :new_c2_f2, ...])

Let's practice!

Julia でのデータ操作