Pipe operator

Description

Pipe operator from magrittr

Usage

lhs %>% rhs

Absorb values into a string containing keys

Description

Populate string templates containing keys with their values. The keys are interpreted as regular expressions. Results can optionally be evaluated as R expressions.

Usage

absorb(
    key, 
    value, 
    text, 
    sep = "_",
    trace = FALSE,
    evaluate = FALSE
)

Arguments

Details

The inputs are iterated in sequential order to replace each pattern with its corresponding value. It is possible that a subsequent pattern could match with a prior result, and hence be replaced more than once. If duplicate keys exist, the placeholder will be filled with a collapsed string of all the values for that key.

Value

• If evaluate = FALSE (default), a character vector the same length as text with all matching patterns replaced by their value.

• Otherwise, a list with the same length as text.

Author(s)

Alex Zajichek

Examples

#Simple example
absorb(
    key = c("mean", "sd", "var"),
    value = c("10", "2", "4"),
    text = 
        c("MEAN: mean, SD: sd",
          "VAR: var = sd^2",
          MEAN = "mean"
        )
)

#Evaluating results
absorb(
    key = c("mean", "mean", "sd", "var"),
    value = c("10", "20", "2", "4"),
    text = c("(mean)/2", "sd^2"),
    sep = "+",
    trace = TRUE,
    evaluate = TRUE
) %>%
    rlang::flatten_dbl()

Find the elements in a list structure that satisfy a predicate

Description

Traverse a list of structure to find the depths and positions of its elements that satisfy a predicate.

Usage

depths(
    list,
    predicate,
    bare = TRUE,
    ...
)
depths_string(
    list,
    predicate,
    bare = TRUE,
    ...
)

Arguments

Details

The input is recursively evaluated to find elements that satisfy predicate, and only proceeds where rlang::is_list when argument bare is FALSE, and rlang::is_bare_list when it is TRUE.

Value

• depths() returns an integer vector indicating the levels that contain elements satisfying the predicate.

• depths_string() returns a character representation of the traversal. Brackets {} are used to indicate the level of the tree, commas to separate element-indices within a level, and the sign of the index to indicate whether the element satisfied predicate (- = yes, + = no).

Author(s)

Alex Zajichek

Examples

#Find depths of data frames
df1 <-
  heart_disease %>%
  
    #Divide the frame into a list
    divide(
      Sex,
      HeartDisease,
      ChestPain
    )

df1 %>%
  
  #Get depths as an integer
  depths(
    predicate = is.data.frame
  )

df1 %>%

  #Get full structure
  depths_string(
    predicate = is.data.frame
  )

#Shallower list
df2 <-
  heart_disease %>%
    divide(
      Sex,
      HeartDisease,
      ChestPain,
      depth = 1
    ) 

df2 %>%
  depths(
    predicate = is.data.frame
  )

df2 %>%
  depths_string(
    predicate = is.data.frame
  )

#Allow for non-bare lists to be traversed
df1 %>%
  depths(
    predicate = is.factor,
    bare = FALSE
  )

#Make uneven list with diverse objects
my_list <-
  list(
    heart_disease,
    list(
      heart_disease
    ),
    1:10,
    list(
      heart_disease$Age,
      list(
        heart_disease
      )
    ),
    glm(
      formula = HeartDisease ~ .,
      data = heart_disease,
      family = "binomial"
    )
  )

#Find the data frames
my_list %>%
  depths(
    predicate = is.data.frame
  )

my_list %>%
  depths_string(
    predicate = is.data.frame
  )

#Go deeper by relaxing bare list argument
my_list %>%
  depths_string(
    predicate = is.data.frame,
    bare = FALSE
  )

Compute descriptive statistics on columns of a data frame

Description

The user can specify an unlimited number of functions to evaluate and the types of data that each set of functions will be applied to (including the default; see "Details").

Usage

descriptives(
    data,
    f_all = NULL,
    f_numeric = NULL,
    numeric_types = "numeric",
    f_categorical = NULL,
    categorical_types = "factor",
    f_other = NULL,
    useNA = c("ifany", "no", "always"),
    round = 2,
    na_string = "(missing)"
)

Arguments

Details

The following fun_key's are available by default for the specified types:

• ALL: length, missing, available, class, unique

• Numeric: mean, sd, min, q1, median, q3, max, iqr, range

• Categorical: count, proportion, percent

Value

A tibble::tibble with the following columns:

• fun_eval: Column types function was applied to

• fun_key: Name of function that was evaluated

• col_ind: Index from input dataset

• col_lab: Label of the column

• val_ind: Index of the value within the function result

• val_lab: Label extracted from the result with names

• val_dbl: Numeric result

• val_chr: Non-numeric result

• val_cbn: Combination of (rounded) numeric and non-numeric values

Author(s)

Alex Zajichek

Examples

#Default
heart_disease %>%
    descriptives()

#Allow logicals as categorical
heart_disease %>%
    descriptives(
        categorical_types = c("logical", "factor")
    ) %>%
    
    #Extract info from the column
    dplyr::filter(
        col_lab == "BloodSugar"
    ) 

#Nothing treated as numeric
heart_disease %>%
    descriptives(
        numeric_types = NULL
    )

#Evaluate a custom function
heart_disease %>%
    descriptives(
        f_numeric = 
            list(
                cv = function(x) sd(x, na.rm = TRUE)/mean(x, na.rm = TRUE)
            )
    ) %>%
    
    #Extract info from the custom function
    dplyr::filter(
        fun_key == "cv"
    ) 

Evaluate a two-argument function with combinations of columns

Description

Split up columns into groups and apply a function to combinations of those columns with control over whether each group is entered as a single data.frame or individual vector's.

Usage

dish(
    data,
    f,
    left,
    right,
    each_left = TRUE,
    each_right = TRUE,
    ...
)

Arguments

Value

A list

Author(s)

Alex Zajichek

Examples

#All variables on both sides
heart_disease %>%
    dplyr::select(
        where(is.numeric)
    ) %>%
    dish(
        f = cor
    )

#Simple regression of each numeric variable on each other variable
heart_disease %>%
    dish(
        f =
            function(y, x) {
                mod <- lm(y ~ x)
                tibble::tibble(
                    Parameter = names(mod$coef),
                    Estimate = mod$coef
                )
            },
        left = where(is.numeric)
    ) %>%
    
    #Bind rows together
    fasten(
        into = c("Outcome", "Predictor")
    )

#Multiple regression of each numeric variable on all others simultaneously
heart_disease %>%
    dish(
        f =
            function(y, x) {
                mod <- lm(y ~ ., data = x)
                tibble::tibble(
                    Parameter = names(mod$coef),
                    Estimate = mod$coef
                )
            },
        left = where(is.numeric),
        each_right = FALSE
    ) %>%
    
    #Bind rows together
    fasten(
        into = "Outcome"
    )

Divide a data frame into a list

Description

Separate a data.frame into a list of any depth by one or more stratification columns whose levels become the names.

Usage

divide(
    data,
    ...,
    depth = Inf,
    remove = TRUE,
    drop = TRUE,
    sep = "|"
)

Arguments

Details

For the depth, use positive integers to move from the root and negative integers to move from the leaves. The maximum (minimum) depth will be used for integers larger (smaller) than such.

Value

A list

Author(s)

Alex Zajichek

Examples

#Unquoted selection
heart_disease %>%
    divide(
        Sex
    )

#Using select helpers
heart_disease %>%
    divide(
        matches("^S")
    )

#Reduced depth
heart_disease %>%
    divide(
        Sex,
        HeartDisease,
        depth = 1
    )
    
#Keep columns in result; change delimiter in names
heart_disease %>%
    divide(
        Sex,
        HeartDisease,
        depth = 1,
        remove = FALSE,
        sep = ","
    )

#Move inward from maximum depth
heart_disease %>%
    divide(
        Sex,
        HeartDisease,
        ChestPain,
        depth = -1
    )

#No depth returns original data (and warning)
heart_disease %>%
    divide(
        Sex,
        depth = 0
    )
heart_disease %>%
    divide(
        Sex,
        HeartDisease,
        depth = -5
    )

#Larger than maximum depth returns maximum depth (default)
heart_disease %>%
    divide(
        Sex,
        depth = 100
    )

Bind a list of data frames back together

Description

Roll up a list of arbitrary depth with data.frame's at the leaves row-wise.

Usage

fasten(
    list,
    into = NULL,
    depth = 0
)

Arguments

Details

Use empty strings "" in the into argument to omit column creation when rows are binded. Use positive integers for the depth to move from the root and negative integers to move from the leaves. The maximum (minimum) depth will be used for integers larger (smaller) than such. The leaves of the input list should be at the same depth.

Value

A tibble::tibble or reduced list

Author(s)

Alex Zajichek

Examples

#Make a divided data frame
list <-
  heart_disease %>%
  divide(
    Sex,
    HeartDisease,
    ChestPain
  )

#Bind without creating names
list %>% 
  fasten

#Bind with names
list %>% 
  fasten(
    into = c("Sex", "HeartDisease", "ChestPain")
  )

#Only retain "Sex"
list %>%
  fasten(
    into = "Sex"
  )

#Only retain "HeartDisease"
list %>%
  fasten(
    into = c("", "HeartDisease")
  )

#Bind up to Sex
list %>%
  fasten(
    into = c("HeartDisease", "ChestPain"),
    depth = 1
  )

#Same thing, but start at the leaves
list %>%
  fasten(
    into = c("HeartDisease", "ChestPain"),
    depth = -2
  )

#Too large of depth returns original list
list %>%
  fasten(
    depth = 100
  )

#Too small of depth goes to 0
list %>%
  fasten(
    depth = -100
  )

Make a kable with a hierarchical header

Description

Create a knitr::kable with a multi-layered (graded) header.

Usage

grable(
    data,
    at,
    sep = "_",
    reverse = FALSE,
    format = c("html", "latex"),
    caption = NULL,
    ...
)

Arguments

Value

A knitr::kable

Author(s)

Alex Zajichek

Heart Disease

Description

This is a cleaned up version of the "heart disease data set" found in the UCI Machine Learning Repository (https://archive.ics.uci.edu/ml/datasets/Heart+Disease), containing a subset of the default variables.

Usage

heart_disease

Format

See "Source" for link to dataset home page

Source

https://archive.ics.uci.edu/ml/datasets/Heart+Disease

Randomly permute some or all columns of a data frame

Description

Shuffle any of the columns of a data.frame to artificially distort relationships.

Usage

muddle(
    data,
    at,
    ...
)

Arguments

Value

A tibble::tibble

Author(s)

Alex Zajichek

Examples

#Set a seed
set.seed(123)

#Default permutes all columns
heart_disease %>%
  muddle

#Permute select columns
heart_disease %>%
  muddle(
    at = c(Age, Sex)
  )

#Using a select helper
heart_disease %>%
  muddle(
    at = matches("^S")
  )

#Pass other arguments
heart_disease %>%
  muddle(
    size = 5,
    replace = TRUE
  )

Is an object one of the specified types?

Description

Check if an object inherits one (or more) of a vector classes.

Usage

some_type(
    object,
    types
)

Arguments

Value

A logical indicator

Author(s)

Alex Zajichek

Examples

#Columns of a data frame
heart_disease %>%
    purrr::map_lgl(
        some_type,
        types = c("numeric", "logical")
    )

Stratify a data frame and apply a function

Description

Split a data.frame by any number of columns and apply a function to subset.

Usage

stratiply(
    data,
    f,
    by,
    ...
)

Arguments

Value

A list

Author(s)

Alex Zajichek

Examples

#Unquoted selection
heart_disease %>%
    stratiply(
        head,
        Sex
    )

#Select helper
heart_disease %>%
    stratiply(
        f = head,
        by = starts_with("S")
    )
    
#Use additional arguments for the function
heart_disease %>%
  stratiply(
        f = glm,
        by = Sex,
        formula = HeartDisease ~ .,
        family = "binomial"
  )

#Use mixed selections to split by desired columns
heart_disease %>%
  stratiply(
        f = glm,
        by = c(Sex, where(is.logical)),
        formula = HeartDisease ~ Age,
        family = "binomial"
  ) 
  

Span keys and values across the columns

Description

Pivot one or more values across the columns by one or more keys

Usage

stretch(
    data,
    key,
    value,
    sep = "_"
)

Arguments

Details

In the case of multiple value's, the labels are always appended to the end of the resulting columns.

Value

A tibble::tibble

Author(s)

Alex Zajichek

Examples

#Make a summary table
set.seed(123)
data <- 
  heart_disease %>%
  dplyr::group_by(
    Sex,
    BloodSugar,
    HeartDisease
  ) %>%
  dplyr::summarise(
    Mean = mean(Age),
    SD = sd(Age),
    .groups = "drop"
  ) %>%
  dplyr::mutate(
    Random =
      rbinom(nrow(.), size = 1, prob = .5) %>%
      factor
  )

data %>%
  stretch(
    key = c(BloodSugar, HeartDisease),
    value = c(Mean, SD, Random)
  )

data %>%
  stretch(
    key = where(is.factor),
    value = where(is.numeric)
  )

data %>%
  stretch(
    key = c(where(is.factor), where(is.logical)),
    value = where(is.numeric)
  )

Evaluate a function on columns conforming to one or more (or no) specified types

Description

Apply a function to columns in a data.frame that inherit one of the specified types.

Usage

typly(
    data,
    f,
    types,
    negated = FALSE,
    ...
)

Arguments

Value

A list

Author(s)

Alex Zajichek

Examples

heart_disease %>%
    
    #Compute means on numeric or logical data
    typly(
        f = mean,
        types = c("numeric", "logical"),
        na.rm = TRUE
    ) 

Compute association statistics between columns of a data frame

Description

Evaluate a list of scalar functions on any number of "response" columns by any number of "predictor" columns

Usage

univariate_associations(
    data,
    f,
    responses,
    predictors
)

Arguments

Value

A tibble::tibble with the response/predictor columns down the rows and the results of the f across the columns. The names of the result columns will be the names provided in f.

Author(s)

Alex Zajichek

Examples

#Make a list of functions to evaluate
f <-
  list(
    
    #Compute a univariate p-value
    `P-value` =
      function(y, x) {
        if(some_type(x, c("factor", "character"))) {
          
          p <- fisher.test(factor(y), factor(x), simulate.p.value = TRUE)$p.value
          
        } else {
          
          p <- kruskal.test(x, factor(y))$p.value
          
        }
        
        ifelse(p < 0.001, "<0.001", as.character(round(p, 2)))
        
      },
    
    #Compute difference in AIC model between null model and one predictor model
    `AIC Difference` =
      function(y, x) {
        
        glm(factor(y)~1, family = "binomial")$aic -
          glm(factor(y)~x, family = "binomial")$aic
        
      }
  )

#Choose a couple binary outcomes
heart_disease %>% 
  univariate_associations(
    f = f,
    responses = c(ExerciseInducedAngina, HeartDisease)
  )

#Use a subset of predictors
heart_disease %>% 
  univariate_associations(
    f = f,
    responses = c(ExerciseInducedAngina, HeartDisease),
    predictors = c(Age, BP)
  )

#Numeric predictors only
heart_disease %>% 
  univariate_associations(
    f = f,
    responses = c(ExerciseInducedAngina, HeartDisease),
    predictors = is.numeric
  )

Create a custom descriptive table for a dataset

Description

Produces a formatted table of univariate summary statistics with options allowing for stratification by one or more variables, computing of custom summary/association statistics, custom string templates for results, etc.

Usage

univariate_table(
    data,
    strata = NULL,
    associations = NULL,
    numeric_summary = c(Summary = "median (q1, q3)"),
    categorical_summary = c(Summary = "count (percent%)"),
    other_summary = NULL,
    all_summary = NULL,
    evaluate = FALSE,
    add_n = FALSE,
    order = NULL,
    labels = NULL,
    levels = NULL,
    format = c("html", "latex", "markdown", "pandoc", "none"),
    variableName = "Variable",
    levelName = "Level",
    sep = "_",
    fill_blanks = "",
    caption = NULL,
    ...
)

Arguments

Value

A table of summary statistics in the specified format. A tibble::tibble is returned if format = "none".

Author(s)

Alex Zajichek

Examples

#Set format
format <- "pandoc"

#Default summary
heart_disease %>%
    univariate_table(
      format = format
    )

#Stratified summary
heart_disease %>%
    univariate_table(
        strata = ~Sex,
        add_n = TRUE,
        format = format
    )

#Row strata with custom summaries with
heart_disease %>%
    univariate_table(
        strata = HeartDisease~1,
        numeric_summary = c(Mean = "mean", Median = "median"),
        categorical_summary = c(`Count (%)` = "count (percent%)"),
        categorical_types = c("factor", "logical"),
        add_n = TRUE,
        format = format
    )