Skip to main content

Apache Spark RDD - Sampling With Replacement and Sampling Without Replacement

Sampling is a popular Spark RDD operation. Sampling With Replacement and sampling without replacement are different ways of doing sampling. This article explains the difference between them.



Sampling with replacement:
Consider a population of potato sacks, each of which has either 12, 13, 14, 15, 16, 17, or 18 potatoes, and all the values are equally likely. Suppose that, in this population, there is exactly one sack with each number. So the whole population has seven sacks. If I sample two with replacement, then I first pick one (say 14). I had a 1/7 probability of choosing that one. Then I replace it. Then I pick another. Every one of them still has 1/7 probability of being chosen. And there are exactly 49 different possibilities here (assuming we distinguish between the first and second.) They are: (12,12), (12,13), (12, 14), (12,15), (12,16), (12,17), (12,18), (13,12), (13,13), (13,14), etc.
Sampling without replacement:
Consider the same population of potato sacks, each of which has either 12, 13, 14, 15, 16, 17, or 18 potatoes, and all the values are equally likely. Suppose that, in this population, there is exactly one sack with each number. So the whole population has seven sacks. If I sample two without replacement, then I first pick one (say 14). I had a 1/7 probability of choosing that one. Then I pick another. At this point, there are only six possibilities: 12, 13, 15, 16, 17, and 18. So there are only 42 different possibilities here (again assuming that we distinguish between the first and the second.) They are: (12,13), (12,14), (12,15), (12,16), (12,17), (12,18), (13,12), (13,14), (13,15), etc.
What’s the Difference?
When we sample with replacement, the two sample values are independent. Practically, this means that what we get on the first one doesn’t affect what we get on the second. Mathematically, this means that the covariance between the two is zero.
In sampling without replacement, the two sample values aren’t independent. Practically, this means that what we got on the for the first one affects what we can get for the second one. Mathematically, this means that the covariance between the two isn’t zero. That complicates the computations. In particular, if we have a SRS (simple random sample) without replacement, from a population with variance , then the covariance of two of the different sample values is , where N is the population size. (A brief summary of some formulas is provided here. For a discussion of this in a textbook for a course at the level of M378K, see the chapter on Survey Sampling in Mathematical Statistics and Data Analysis by John A. Rice, published by Wadsworth & Brooks/Cole Publishers. There is an outline of an slick, simple, interesting, but indirect, proof in the problems at the end of the chapter.)
Population size — Leading to a discussion of “infinite” populations.
When we sample without replacement, and get a non-zero covariance, the covariance depends on the population size. If the population is very large, this covariance is very close to zero. In that case, sampling with replacement isn’t much different from sampling without replacement. In some discussions, people describe this difference as sampling from an infinite population (sampling with replacement) versus sampling from a finite population (without replacement).

Comments

Popular posts from this blog

Basics Of Hadoop Ecosystem - 2

Basics of Hadoop Ecosystem – 2 Apache Ambari is a web-based tool for provisioning, managing, and monitoring Apache Hadoop clusters. It has a very simple yet highly interactive UI to install various tools and perform various management, configuration and monitoring tasks.  Introduction: In part 2 of the blog series I will cover the other core components of a Hadoop Framework including querying part, external integration, data exchange, co-ordination and managing as well as monitoring of Hadoop Clusters. Please refer to Basic’s of Hadoop Ecosystem  Part 1 ... QUERYING: Pig seems quite useful however; I am more of an SQL person. For those of us who still like SQL, we have SQL for Hadoop. HIVE Hive is a distributed data warehouse built on atop of HDFS to manage and organize large amounts of data. It provides a query based on SQL semantics  (HiveQL)  which is translated by the runtime engine to Map Reduce jobs for querying the data. Sche...

Data Engineering - Tools & Intro

Data Engineering - Tools & Intro So I just realized that I am here after a month or so. I was busy at work and traveling. I am starting a kind of new series, I say it Data Engineering Series in which I will be discussing different tools. Of course, I am not able to discuss the entire concept of Data Engineering neither I know it as I will be learning myself. What is Data Engineering? Data Engineering is all about developing, maintaining systems that are responsible for transferring data in large volumes and make it available for analysts and data scientists to use it for analyzing and data modeling. Data engineering is a superset of Data Science or the subset, not clear to me but the collaboration of data engineers and scientists fruits useful data-driven solutions. Data Engineering tools It consists of several tools. Some are dealing with data storage while others with analysis and ETL. Ofcourse, Apache Kafka is one of them. The others tools that I might be covering are Apache...

Let's Understand Ten Machine Learning Algorithms

Ten Machine Learning Algorithms to Learn Machine Learning Practitioners have different personalities. While some of them are “I am an expert in X and X can train on any type of data”, where X = some algorithm, some others are “Right tool for the right job people”. A lot of them also subscribe to “Jack of all trades. Master of one” strategy, where they have one area of deep expertise and know slightly about different fields of Machine Learning. That said, no one can deny the fact that as practicing Data Scientists, we will have to know basics of some common machine learning algorithms, which would help us engage with a new-domain problem we come across. This is a whirlwind tour of common machine learning algorithms and quick resources about them which can help you get started on them. 1. Principal Component Analysis(PCA)/SVD PCA is an unsupervised method to understand global properties of a dataset consisting of vectors. Covariance Matrix of data points is analyzed here to un...