Constant Fill

Constant fill is a useful method of imputation when the missing, NA values in a dataset can be assumed to be a certain value. Sometimes when moving data between different platforms and software, certain values in the data may be 'lost' or flagged as NA erroneously. Using constant fill imputation lets you replace these missing values with a default.

In this exercise, you'll impute missing values for the car_sales time series, which represents daily car sales for an employee at a car dealership.

This exercise is part of the course

Manipulating Time Series Data in R

Exercise instructions

Determine the total number of NA values in the car_sales time series.
Use constant fill imputation to fill the missing values of car_sales with 0; assign this to the car_sales_filled variable.
Autoplot the car_sales_filled time series.

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Add together the number of NA values
___

# Fill in missing values with zero
___ <- ___

# Autoplot the filled time series
autoplot(___) + 
  labs(y = "Daily Car Sales") + 
  theme_light()

Edit and Run Code

This exercise is part of the course

Manipulating Time Series Data in R

IntermediateSkill Level

4.9+

Start Course for Free

You'll learn the foundations of what a time series represents, how to retrieve summary statistics about the data in a time series, and how to visually interpret a time series plot as part of the exploration step of your analysis. You’ll also cover how to manage date and time information within R objects and ways of incorporating consistent formatting for dates.

Exercise 1: What is time series data?Exercise 2: Plotting time series with autoplot Exercise 3: Time series summary Exercise 4: Interpreting time series plots Exercise 5: Temporal data classes in R Exercise 6: Finding the class Exercise 7: is-dot functions Exercise 8: lubridate's as_date Exercise 9: Formatting dates in R Exercise 10: Conversion specifications Exercise 11: Parsing and formatting dates

Here, you’ll learn to retrieve key attributes of time series information, such as the range in time of the data and how often the data were sampled, to understand your data better. You'll also be introduced to the zoo package, which contains tools and functions for creating and manipulating time series objects. Many data science applications in R use the data frame paradigm; you'll learn how to convert between a data frame and a time series.

Exercise 1: Time series attributes Exercise 2: Temporal attributes and decimal dates Exercise 3: Comparing temporal attributes Exercise 4: Time series objects with the zoo package Exercise 5: Creating a zoo Exercise 6: zoo and ggplot2 Exercise 7: Manipulating zoo objects Exercise 8: Updating and replacing indices Exercise 9: Finding overlapping indices Exercise 10: Converting between zoo and data frame Exercise 11: Moving between data.frame and zoo

You’ll cover how to subset a window from a time series to focus on a particular period of interest. You’ll see that when working with real-world time series data, the timespan of your dataset may cover more information than you need, which can clutter your visualizations. You’ll sample time series data at various rates, such as every minute, hour, month, or year. Further, by aggregating your data, you’ll learn to see the overall trends in the data using the xts package. You'll also learn methods of 'imputing' your data – filling in missing values with constant fill, LOCF, or linear interpolation methods.

Exercise 1: Subsetting a window of observations Exercise 2: Selecting a window from a time series Exercise 3: Finding an observation from a specific date Exercise 4: Logical expressions and subsets Exercise 5: Monthly and quarterly data Exercise 6: zoo::yearmon Exercise 7: zoo::yearqtr Exercise 8: Resampling and aggregating observations Exercise 9: Aggregating data Exercise 10: Custom aggregation level Exercise 11: Plotting an aggregated time series with ggplot2 Exercise 12: Imputing missing values Exercise 13: Imputation methods Exercise 14: Constant Fill

Current Exercise

Exercise 15: Last observation carried forward (LOCF)Exercise 16: Linear interpolation

You’ll learn to create “rolling” windows of a time series that move, or "roll" along with data, making it possible to summarize trends in the data across time, such as the average over success months of observations or the sum over several weeks of sales. Overall summary statistics, like mean, median, sum, maximum, and so on, do not always provide insight into how data changes over time, and rolling windows will allow you to compute statistics dynamically. In addition to rolling windows, you will also learn how to create expanding windows, which show how these summary statistics approach their final value.

Exercise 1: What is a rolling window?Exercise 2: Rolling window functions Exercise 3: Rolling windows versus aggregation Exercise 4: Applying functions to rolling windows Exercise 5: Rolling minimum Exercise 6: Rolling apply with a custom function Exercise 7: Expanding windows Exercise 8: Rolling versus expanding windows Exercise 9: Expanding sum Exercise 10: Expanding mean Exercise 11: Congratulations!