In UBC-MDS/hamr: Handle All Missing Values

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Introduction

Our package intends to explore the pattern of missing values in users' dataset, imputes the missing values using three simple methods and compare the results of different methods.

We found Amelia and vis_dat packages that are similar but only visualize the missing data. We thought this would be better package for users who do not have much experience in data wrangling.

vis_missing()

Introduction

Explore the pattern of missing values in a dataset.

Function

vis_missing(df, colour="default", missing_val_char = NA)

Parameters:

• df: a data frame or matrix with missing values
• the original dataset with missing values
• colour: string
• colour scheme for plotting - deprecated
• missing_val_char:
• the missing value character in the data frame, one of NA, "", " ", "?"

Return:

• a heatmap plot

Example

dfm <- data.frame(x = c(1, 2, 3), y = c(0, 10, NaN))
vis_missing(dfm, "", NaN)

impute_missing()

Introduction:

imputes the missing values in a specified column with three simple methods: complete case, mean imputation and median imputation

Function

impute_missing(dfm, col, method, missing_val_char)

Parameters:

• dfm: a data frame or a numerical matrix

• the original dataset with missing values

• col: string

• a column name

• method: string

• a method name, expected one of "CC", "MIP", "DIP"

• missing_val_char:

• the format of the missing value, expected one of NA, NaN, "" and "?"

Return:

• a data frame with no missing values in the specified column

Example

dfm <- data.frame(x = c(1, 2, 3), y = c(0, 10, NaN))
impute_missing(dfm, "y", "MIP", NaN)

compare_model()

Introduction:

Compare the results of different methods. * This function will call function impute_missing() for several methods and return a table with some statistical information of the specified feature before and after imputation of different methods

Function

compare_model()

Parameters:

• df (ndarray): -- the original dataset with missing values

• Feature (str): -- Name of a specified feature from the original dataset

• containing missing values that needs to be imputed.

• methods (str or list): -- the methods that users want to compare (default: ["CC","IMP"])

• Supporting methods are: CC - Complete Case MIP - Imputation with mean value DIP - Imputation with median value
• missing_val_char (str): -- missing value types.
• Supporting types are: "NaN" - Not a Number "" - Blank "?" - Question mark

Return:

• a summary table comparing the summary statistics: count, mean, std, min, 25%, 50%, 75%, max.

Example

dfm <- data.frame(x = c(1, 2, 3), y = c(0, 10, NaN))
compare_model(dfm, "y", "MIP", "NaN")

UBC-MDS/hamr documentation built on May 25, 2019, 1:36 p.m.