manymodelr 0.2.1

A look at some of the key functions.

• agg_by_group

As one can guess from the name, this function provides an easy way to manipulate grouped data. We can for instance find the number of observations in the iris data set. The formula takes the form x~y where y is the grouping variable(in this case Species). One can supply a formula as shown next.

agg_by_group(iris,.~Species,length)
#> Grouped By[1]:    Species 
#> 
#>      Species Sepal.Length Sepal.Width Petal.Length Petal.Width
#> 1     setosa           50          50           50          50
#> 2 versicolor           50          50           50          50
#> 3  virginica           50          50           50          50

head(agg_by_group(mtcars,cyl~hp+vs,sum))
#> Grouped By[2]:    hp  vs 
#> 
#>    hp vs cyl
#> 1  91  0   4
#> 2 110  0  12
#> 3 150  0  16
#> 4 175  0  22
#> 5 180  0  24
#> 6 205  0   8

• multi_model_1

This is one of the core functions of the package. Since the function uses caret backend, we need to load the function before we can use it. To avoid, several messages showing up, we use the function suppressMessages. This assumes that one is familiar with machine learning basics. We specify our model types and we use the argument valid=TRUE to specify that we are dealing with validation. Had we wanted to predict on unseen test data, then this argument would be set to FALSE.

suppressMessages(library(caret))
train_set<-createDataPartition(iris$Species,p=0.8,list=FALSE)
valid_set<-iris[-train_set,]
train_set<-iris[train_set,]
ctrl<-trainControl(method="cv",number=5)
set.seed(233)
 m<-multi_model_1(train_set,"Species",".",c("knn","rpart"),
"Accuracy",ctrl,newdata =valid_set,valid=TRUE)
#> [1] "Returning Metrics"

The above message tells us that the model has returned our metrics for each of the model types we specified. These can be extracted as shown below. Other return values include predictions and a summary of the model.

m$Metrics
#> # A tibble: 1 x 2
#>     knn rpart
#>   <dbl> <dbl>
#> 1 0.933 0.967

head(m$Predictions)
#> # A tibble: 6 x 2
#>   knn    rpart 
#>   <fct>  <fct> 
#> 1 setosa setosa
#> 2 setosa setosa
#> 3 setosa setosa
#> 4 setosa setosa
#> 5 setosa setosa
#> 6 setosa setosa

• modeleR

Yet another core function, this allows us to (currently) conveniently perform linear regression and analysis of variance. Here we are simultaneously building and using our model to predict. This is particularly useful if you already know how well a given model models your data. We can extract the results just like we did above.

iris1<-iris[1:60,]
iris2<-iris[60:nrow(iris),]
m1<-modeleR(iris1,Sepal.Length,Petal.Length,
        lm,na.rm=TRUE,iris2)
head(m1$Predictions)
#>    Predicted
#> 60  5.985141
#> 61  5.821972
#> 62  6.107518
#> 63  6.025933
#> 64  6.311478
#> 65  5.862764

• get_var_cor

As the name suggests, this function is useful when carrying out correlation tests as shown below. Setting get_all to TRUE implies that all the variables are correlated(from exploratory data analysis) and you just want to see what the correlation(s) is(are). The variant function get_var_corr_ (note the underscore at the end provides a convenient way to get correlations for combinations of variables(pairs).)

get_var_corr(mtcars, "mpg",get_all = TRUE)
#>    Comparison_Var Other_Var      p_value Correlation    lower_ci
#> 1             mpg       cyl 6.112687e-10  -0.8521620 -0.92576936
#> 2             mpg      disp 9.380327e-10  -0.8475514 -0.92335937
#> 3             mpg        hp 1.787835e-07  -0.7761684 -0.88526861
#> 4             mpg      drat 1.776240e-05   0.6811719  0.43604838
#> 5             mpg        wt 1.293959e-10  -0.8676594 -0.93382641
#> 6             mpg      qsec 1.708199e-02   0.4186840  0.08195487
#> 7             mpg        vs 3.415937e-05   0.6640389  0.41036301
#> 8             mpg        am 2.850207e-04   0.5998324  0.31755830
#> 9             mpg      gear 5.400948e-03   0.4802848  0.15806177
#> 10            mpg      carb 1.084446e-03  -0.5509251 -0.75464796
#>      upper_ci
#> 1  -0.7163171
#> 2  -0.7081376
#> 3  -0.5860994
#> 4   0.8322010
#> 5  -0.7440872
#> 6   0.6696186
#> 7   0.8223262
#> 8   0.7844520
#> 9   0.7100628
#> 10 -0.2503183

To get correlations for only select variables, one could work as follows:

get_var_corr(mtcars,comparison_var = "cyl",
             other_vars = c("disp","mpg"),get_all = FALSE)
#>   Comparison_Var Other_Var      p.value Correlation   lower_ci   upper_ci
#> 1            cyl      disp 1.802838e-12   0.9020329  0.8072442  0.9514607
#> 2            cyl       mpg 6.112687e-10  -0.8521620 -0.9257694 -0.7163171

Similarly, get_var_corr_ (note the underscore at the end) provides a convenient way to get combination-wise correlations.

head(get_var_corr_(mtcars),6)
#>   Comparison_Var Other_Var      p.value Correlation    lower_ci   upper_ci
#> 1            mpg       cyl 6.112687e-10  -0.8521620 -0.92576936 -0.7163171
#> 2            mpg      disp 9.380327e-10  -0.8475514 -0.92335937 -0.7081376
#> 3            mpg        hp 1.787835e-07  -0.7761684 -0.88526861 -0.5860994
#> 4            mpg      drat 1.776240e-05   0.6811719  0.43604838  0.8322010
#> 5            mpg        wt 1.293959e-10  -0.8676594 -0.93382641 -0.7440872
#> 6            mpg      qsec 1.708199e-02   0.4186840  0.08195487  0.6696186

• rowdiff

This is useful when trying to find differences between rows. The direction argument specifies how the subtractions are made while the exclude argument is currently used to remove non-numeric data. Using direction="reverse" performs a subtraction akin to x-(x-1) where x is the row number.

head(rowdiff(iris,exclude = "non_numeric",direction = "reverse"))
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width
#> 1           NA          NA           NA          NA
#> 2         -0.2        -0.5          0.0         0.0
#> 3         -0.2         0.2         -0.1         0.0
#> 4         -0.1        -0.1          0.2         0.0
#> 5          0.4         0.5         -0.1         0.0
#> 6          0.4         0.3          0.3         0.2