Sampling Theory Course

Basics

• The Sampling Theory course is taught by Ha-Young Shin at Soongsil University
• The course is in Spring 2025
• Classes are held on Wednesdays from 15:00-16:15 in 진리관 516 and Thursdays from 13:30-14:45 in 진리관 522
• The instructor can be reached by email at [email protected]
• The office is located in 베어드홀 508
• Office hours are Mondays and Tuesdays from 14:30-16:30, or by appointment
• The textbook for the course is Elementary Survey Sampling, 7th edition by Scheaffe et al., available as a PDF online
• The course will focus on theory

Language and Evaluation

• Lectures, notes, quizzes, and exams will be in English
• Office hours may be conducted in Korean or English
• Students may ask questions in Korean
• Students should ask the instructor to speak slower if needed
• Students should ask the instructor to clarify anything that is unclear
• Attendance is 5% of the final grade
• Quizzes are 25% of the final grade
• The midterm exam is 35% of the final grade
• The final exam is 35% of the final grade
• Optional homework problems may be suggested
• Attendance is a small part of the overall grade
• Absences will be excused for legitimate reasons
• More than 10 absences will result in a failing grade for the course

Quizzes and Exams

• There will be 6 quizzes, each lasting 10-15 minutes
• The lowest quiz score will be dropped when calculating the final grade
• Quizzes will be held at the beginning of class
• Quizzes will cover material up to and including the previous Thursday
• The first quiz is on Wednesday, March 26
• The midterm exam is on Wednesday, April 23, during class time
• The final exam is on Thursday, June 12, during class time
• Exam dates are subject to change
• Each exam will cover material up to and including the previous Thursday

Course Outline

• The course outline is subject to change
• The course will cover chapters 2–10 in the textbook
• Introduction and basic concepts will be covered in 2 lectures
• Simplified random sampling requires 4 lectures
• Stratified random sampling requires 4 lectures
• Ratio, regression and difference estimation requires 4 lectures
• Systematic sampling requires 3 lectures
• There will be a midterm exam
• Cluster sampling requires 4 lectures
• Two-stage cluster sampling requires 4 lectures
• Estimating the population size requires 2 lectures
• There will be a final exam

Survey Terminology

• Element: An object on which a measurement is taken
  • Example: In an election survey, the element is a registered voter and the measurement is their intended vote
• Population: A collection of elements about which we wish to make an inference
  • Example: In an election survey, the population is all people who will vote in the election
• Parameter: Some property based on measurements of the population
  • This is what we are trying to infer using the sample
• Sampling unit: Nonoverlapping collections of elements from the population that cover the entire population
  • Often the sampling unit is the element, but not always. For example, entire households could be sampled then measurements taken on individual members; see cluster sampling
• Sample: The collection of elements in a collection of sampling units drawn from the population
• Statistic/estimator: An estimate of the parameter calculated based on measurements of the sample
• (Sampling) frame: A collection of sampling units from which the sample is drawn
  • The sampling frame may not equal the entire population of interest, but it is the best available option. For example, in an election survey, the frame could be the list of registered voters
• Survey: The measurement of elements in a sample

Why Conduct Surveys?

• A census involves taking a measurement for every element in a population
• Censuses are often impractical due to time, money, and effort constraints
• Surveys, combined with statistics/probability, enable inference about a population using samples
• Relationship: sample relates to population; survey relates to census; statistic relates to parameter
• Sampling theory allows both statistical inference after a survey and enables the design of surveys themselves
  • For example, determining the necessary sample size to achieve a 2% margin of error
• Two main methods exist for taking measurements on a sample in statistics
  • Experiments: Action is performed on the elements before the measurement is taken
  • Observational studies: Elements are merely observed/measured
  • Usually the word 'survey' means observational study

Errors in Surveys

• Elements must be drawn randomly from the population to ensure representativeness
• Key to all surveying is representativeness which is important, as the statistic is usually not identical to the parameter
• Non-sampling error: Results from non-representativeness, which can be attributed to:
  • Selection bias: Is the group of people initially contacted representative? For example, contacting people via landlines/cellphones?
  • Nonresponse: People can not be forced to respond to surveys. Even if the group of people initially contacted is representative, is the group of respondents?
  • Inaccurate response: For example, respondents changing their mind after the survey and before the election, or lying about their vote
• Non-sampling error is the most difficult problem to deal with in surveying and better data collection methods are required
• Sampling error: Results from randomness since the sample is only a subset of the population
• Sampling error can be addressed and quantified using our knowledge of statistics/probability