Sampling is a process of selecting part of target population to be studied. In an optimal world, we would measure everybody. Measuring everybody would allow us to perfectly describe the population and get to something called “True mean”. If we get everybody’s height, we don’t need to run statistics. We don’t need to “test” for group differences, we can be certain about our assertions regardless of statistics.
But we are not in an wishful and perfect world adn we only can measure a fixed amout of people - only a hundred inhabitants of France. How then we pick those that perfectly represent the entire population of France? Entire population of Young adults?Entire population of home schooled children?
| Research question | Population | Ideal practical sample | Real sample |
|---|---|---|---|
| How many faces can human memory store? | Everybody | 200 random people from around the world | 60 people from PSYCH 101 class |
| Is the new antipsychotic drug better than the previous one? | All patients on the previous drug/all patients | 100 random patients from countries where the drug is going to be sold | 15 patients from John Hopkins or any other hospital willing to participate |
| Do the new contraception pills affect cognitive skills? | All women on contraceptives | 200 random women on contraceptives | 30 women who visit gynaecologist who is being paid to do the study |
| How do babies interact with their peers? | All babies between 0 to 3 years | 50 observations in different situations and settings | 10 observations done in a kindergarten close to the research lab |
Sampling bias = not selecting a sample representative of the population
Time, money and personnel troubles arise in random samples, biases in nonprobability samples
It is not clearly stated why a sample was selected in this way
The research conclusion does not reflect sampling bias –Research claims that all students are lazy, but only tested students in two classes in three high schools in Virginia
##Sampling process
Probability methods
Non probability methods
Every person in target population has an equal chance of ending in the sample. Usually done by making a list of all the people in the population and then randomly selecting N names.
| Pros | Cons |
|---|---|
| Statistically the best solution, Least troubles defending it before other scientists | Where do you get the list of people to randomize from?, Nearly impossible in most cases, Time and money issues, Large decay of sample (just selecting people does not mean they will participate), Not necessarily the best in all cases, People usually refuse or they feel their anonymity was breached (where did you get my contact??) |
Rather than selecting each individual in target population, we focus on clusters where these people occur in larger groups (workplace, school, clubs etc.). Every group then has an equal chance of ending in a sample. Therefore we select a four football clubs and text all their members (N = 110), rather than selecting 110 players from across the country.
| Pros | Cons |
|---|---|
| Much cheaper, Fast – people come to clubs/schools on their own volition, Focused groups | Very biased (we can select the best group or the worst), The bias is usually multiplied by the number of people in the cluster (30 bright students, vs 30 average students), Needs permission of the entire club/class/school |
Selecting random sample from within subgroups defined by the research aim. You want to draw a stratified sample of 50 children from a school containing 75% girls and 25% boys-> girls need to be three times as likely to be in the sample.
| Pros | Cons |
|---|---|
| Very targeted and focused on the question, Allows for smaller samples while maintaining validity and statistical power, Allows for controlling of extraneous variables | You need to know the subgroups and extraneous variables before the research takes place, sometimes this is not achievable, it is easier than to just give people questionnaire and control for variables using statistics |
People simply apply to participate in the research for university credit, money or just out of curiosity. Or researchers ask friends, relatives etc. They usually need to meet certain criteria (e.g. males 25-36 in a relationship; students of visual arts; people working in an international company etc.)
| Pros | Cons |
|---|---|
| Fast and convenient, People usually come to you, Universities promote research, Sometimes it just doesn’t matter, who participates, so why make it hard? | Extremely biased in most cases, Cannot be easily generalised to population, Somewhat unethical |
Sampling is a process of selecting people from target population Controlling is a process of eliminating extraneous variables
| Name | URL |
|---|---|
| Randomising samples out of different distributions | https://github.com/hejtmy/porgR/blob/master/sampling/sampling.md |