R Tutorial: Data Synthesis

Want to learn more? Take the full course at https://learn.datacamp.com/courses/data-privacy-and-anonymization-in-r at your own pace. More than a video, you'll learn hands-on coding & quickly apply skills to your daily work. --- Now, that you've learned how to remove personally identifiable information and alter data to provide more privacy, you'll learn how to create synthetic or fake data sets by sampling from probability distributions. When data resembles a probability distribution, we can use that distribution to generate new data. This is one approach to achieving anonymization, because the idea is we are creating a fake data set with fake people that is still statistically representative of the original data. For this lesson, you'll be looking at the Male Fertility data. We see that there are 100 observations and several variables that contain personal information. Suppose you want to release participant information on Childhood Disease. You can generate a synthetic data set by sampling from a binomial distribution. In order to sample from a binomial distribution, you must know the proportion of participants who had a Childhood Disease. To do this, we use summarise_at() to select the variable Child_Disease and find the average using the mean() function. You see that 87 percent of the participants experienced a childhood disease. You can now sample from a binomial distribution by using the function rbinom(). You will generate 100 samples, set the size to 1, and use 87% as your probability. From this, you can create a synthetic data set where the proportion of participants who had a Childhood Disease is now 83%. You can generate other types of data as long as you identify a proper probability distribution. Here is a histogram of the values from Hours Sitting that you saw earlier in the video. The histogram roughly resembles a normal distribution, but has values that are skewed to the right. So, you should apply a log transformation. Also, note that the v