README.md

In korydjohnson/rai: Revisiting-Alpha-Investing for Polynomial Regression

rai

Overview

rai provides a modified implementation of stepwise regression that greedily searches the space of interactions among features in order to build polynomial regression models. Furthermore, the hypothesis tests conducted are valid post model selection due to the use of a revisiting procedure that implements an alpha-investing rule. As a result, the set of rejected sequential hypotheses is proven to control the marginal false discover rate. When not searching for polynomials, the package provides a statistically valid algorithm to run and terminate stepwise regression.

For more information, see the corresponding paper: Revisiting Alpha-Investing: Conditionally Valid Stepwise Regression.

Installation

You can install the released version of rai from CRAN with:

install.packages("rai")

And the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("korydjohnson/rai")

Usage

library(rai)
data("CO2")
theResponse = CO2$uptake
theData = CO2[ ,-5]
rai_out = rai(theData, theResponse)

The returned object includes a linear model object of the identified model:

summary(rai_out$model)
#> 
#> Call:
#> lm(formula = aucOut$formula, data = data.frame(y = theResponse, 
#>     aucOut$subData))
#> 
#> Residuals:
#>     Min      1Q  Median      3Q     Max 
#> -8.3982 -2.4690  0.1839  2.6110  9.9281 
#> 
#> Coefficients:
#>                                                Estimate Std. Error t value
#> (Intercept)                                   1.354e+01  1.969e+00   6.878
#> I(TypeMississippi)                           -2.571e+00  2.291e+00  -1.122
#> I(conc)                                       8.830e-02  8.722e-03  10.123
#> I(conc * conc)                               -6.007e-05  7.794e-06  -7.707
#> I(Treatmentchilled)                          -3.692e+00  1.094e+00  -3.374
#> I(PlantMc2)                                  -5.591e+00  1.812e+00  -3.086
#> I(conc * TypeMississippi)                    -2.495e-02  7.819e-03  -3.191
#> I(conc * Treatmentchilled * TypeMississippi) -1.028e-02  3.010e-03  -3.414
#> I(conc * conc * conc * TypeMississippi)       1.839e-08  6.702e-09   2.744
#>                                              Pr(>|t|)    
#> (Intercept)                                  1.57e-09 ***
#> I(TypeMississippi)                            0.26526    
#> I(conc)                                      1.12e-15 ***
#> I(conc * conc)                               4.31e-11 ***
#> I(Treatmentchilled)                           0.00118 ** 
#> I(PlantMc2)                                   0.00285 ** 
#> I(conc * TypeMississippi)                     0.00207 ** 
#> I(conc * Treatmentchilled * TypeMississippi)  0.00104 ** 
#> I(conc * conc * conc * TypeMississippi)       0.00758 ** 
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 4.029 on 75 degrees of freedom
#> Multiple R-squared:  0.8746, Adjusted R-squared:  0.8612 
#> F-statistic: 65.36 on 8 and 75 DF,  p-value: < 2.2e-16

You can view a summary of the series of tests conducted by rai and the results of those tests by calling summary on the returned object:

summary(rai_out)
#> $plot_rS

#> 
#> $plot_wealth

#> 
#> $experts
#> # A tibble: 9 x 4
#>   expert          nRej nFeatures order
#>   <chr>          <int>     <int> <dbl>
#> 1 SMarginal          4        15     1
#> 2 SPoly 14           0         1     2
#> 3 SPoly 1            2         2     3
#> 4 SPoly 1_1          0         3     4
#> 5 SPoly 15           0         4     5
#> 6 SPoly 11           0         5     6
#> 7 SPoly 1_14         2         6     7
#> 8 SPoly 1_14_15      0         7     8
#> 9 SPoly 1_1_1_14     0         8     9
#> 
#> $tests
#> # A tibble: 8 x 4
#>   timesTested count nExperts expert        
#>         <int> <int>    <int> <chr>         
#> 1          22     1        1 SPoly 1_14_15 
#> 2          12    15        3 SPoly 11      
#> 3          11    10        5 SPoly 1_1_1_14
#> 4          10     3        3 SPoly 1_14    
#> 5           7     1        1 SMarginal     
#> 6           6     1        1 SPoly 1       
#> 7           5     2        1 SMarginal     
#> 8           1    16        7 SPoly 1_1     
#> 
#> $epochs
#> # A tibble: 12 x 3
#>    epoch  rCrit  nRej
#>    <dbl>  <dbl> <int>
#>  1     1 0.8        0
#>  2     2 0.64       0
#>  3     3 0.512      0
#>  4     4 0.410      0
#>  5     5 0.328      2
#>  6     6 0.262      1
#>  7     7 0.210      1
#>  8     8 0.168      0
#>  9     9 0.134      0
#> 10    10 0.107      3
#> 11    11 0.0859     1
#> 12    12 0.0687     0
#> 
#> $stats
#> $stats$maxPotentialIncrease_raiPlus
#> [1] 0.01077552
#> 
#> $stats$nTests
#> [1] 381
#> 
#> $stats$nEpochs
#> [1] 12
#> 
#> $stats$nFeatures
#> [1] 8
#> 
#> $stats$poly
#> $stats$poly$tableDegrees
#>   degree Freq
#> 1      1    4
#> 2      2    2
#> 3      3    1
#> 4      4    1
#> 
#> $stats$poly$tableInteraction
#>   nUniqueFeatures Freq
#> 1               1    5
#> 2               2    2
#> 3               3    1
#> 
#> 
#> $stats$rS
#> [1] 0.8745565
#> 
#> $stats$nFeaturesTested
#> [1] 49
#> 
#> $stats$nHypothesisTests
#> [1] 150

Necessary functions are provided to use rai within a caret workflow:

# fitControl <- caret::trainControl(method = "repeatedcv",
#                            number = 5, ## 5-fold CV...
#                            repeats = 5)  ## repeated 5 times
# caret::train(x=theData, y=theResponse, method=rai_caret, trControl = fitControl)

korydjohnson/rai documentation built on April 21, 2022, 2:59 p.m.