Validity

For a study to be valid, the conclusions from the study need to represent what or who they claim to represent.  The sample needs to represent the population.  Is there room for samples that do not represent the population?  Yes, but the claims that we can make from those samples are limited.

Sampling Techniques

The sampling techniques can be characterized by techniques that are random, random from groups and non-random.  In random sampling techniques, there is an equal probability of each item in the population being chosen for the sample.

Random Sampling Techniques

Simple Random Sample:  A simple random sample is where a percentage of the population are randomly chosen.  A technique to draw a random individual is used, such as a random number generator.  

Systematic Random Sample:  The population is sampled from an ordered list which chooses individuals at regular intervals.  For example, if the population has a size of 100, and you wanted to sample 20 of them, you would divide 100 by 20 which would give you twenty intervals of five. You would randomly select one of the first five people, and then take every fifth person next on the ordered list.

Random Sampling from Groups

Stratified Random Sample:  The population is divided into groups based on certain demographics. (definition:Traits that describe the characteristics of a population. Characteristics could include age, income, education, gender, race, etc.) The same percentage of each group is selected at random using the same technique as a simple random sample, but within each group. This would allow each group to be represented in the same proportion in the sample as they are in the population.  For example, if you had 3 stratas, one with 50 people, one with 100 people, one with 200 people and you wanted 10%, you would take 5, 10 and 20 people from the groups respectively.

Cluster Random Sample:  The population is divided, usually geographically, into groups that generally have the same size.  A certain number of groups are randomly chosen, and every individual in the chosen groups are chosen for the sample.

Multi-Stage Random Sample:  The population is divided into a hierarchy and a random sample is chosen at each level.  For example, if you wanted to sample a city, you may randomly choose some subdivisions or buildings, and randomly choose some streets or floors, and randomly choose some addresses from each.  

Non-Random Sampling

In non-random sampling, each member of the population does not have the same probability of being chosen.

Voluntary Sample:  A voluntary sample is where participants choose whether or not they will participate.  Sometimes, the participants contact the researchers after being offered some kind of incentive.  A common example of a voluntary sample is an online survey.

Convenience Sample:  Individuals are chosen from a population who are easy to access.  An example of this would be standing outside your school cafeteria and giving a survey to people as they walk by.

Work through the Sampling Techniques Interactive tool to help visualize the different techniques.

SamplingTechniques

Long Description

Bias in Sampling

There are two types of bias that may affect the validity and reliability of a study or survey.  You were introduced to these in a previous activity.  The first is a more general type of bias, sampling bias.  The second is a more specific type, non-response bias.

Sampling Bias

Definition:  A sample is biased when it does not represent the population.  This may be because the number of people surveyed was too low.  However, you saw in Activity 3, the example of the study in May of 2017, where the Angus Reid Institute released the results of an online poll regarding purchasing tickets to live events like concerts.  They polled 1517 Canadian adults, who are members of the Angus Reid Forum.  

Non-response Bias

Definition:  Non-Response Bias is a type of sampling bias where the sample does not accurately represent the population because a particular important segment of the population is missing.  The non-respondents are different than the respondents in meaningful ways.

This is a specific type of bias because although the sample does not represent the population, it can often be fixed by focusing on retrieving the segment of the population that is missing.  For example, if you want to survey the population of your school, there are certain circumstances where you may leave out a particular segment of the population that may not be present the day you survey.  Examples of reasons why people may be missing are: co-op students can be at placements, Grade 9's could be away at orientation, senior students might be away on a field trip, or there may be students enrolled in an off-site alternate education plan.