Table 3 The quantiles of the minimum selection sizes of the four approaches for the Simulation Examples 2–4.

From: Functional random forests for curve response

Quantiles of \(M^\text{{{a}}} \)

Method

\(5\%\)

\(25\%\)

\(50\%\)

\(75\%\)

\(95\%\)

Simulation 2

FunFor

2.00

2.00

2.00

3.00

4.00

FRF

2.00

2.00

4.00

9.00

18.00

Splinetree

16.00

19.00

22.00

24.00

28.00

Refund

2.00

2.00

2.00

2.00

2.00

Simulation 3

FunFor

3.00

4.00

5.00

6.00

7.05

FRF

2.0

2.0

2.0

3.0

5.1

Splinetree

32.00

35.00

38.00

41.00

46.05

Refund

2.00

2.00

2.00

2.00

2.00

Simulation 4

FunFor

2.00

2.00

3.00

3.00

5.05

FRF

3.00

12.00

22.50

36.25

62.05

Splinetree

16.00

38.00

58.00

81.00

94.00

Refund

2.00

2.00

2.00

98.00

100.00

  1. \(^\text{{{a}}}\)M stands for the minimum selection size that is required to include all of the true predictors in each replication. The closer of different quantiles, the robuster.