Sampling Flashcards
What does the term target population mean?
The population for which your results may be of benefit
Population for whom your research results may improve NHS care
What does the term sampling mean?
The proportion of a given population chosen to study
What is probability sampling?
Every member of chosen population has a known probability of being included in the study
What is non probability sampling?
Not random, used when probability sampling is not practical or possible
What is random sampling?
Probability
Every individual has an equal chance of being selected
Good for: generalisation
Bad for: representation
What is stratified sampling?
Probability
Divide population into sub groups then take random sample as above
Good for: representing minority subgroups - representative
What is systematic sampling?
Probability
A method of choosing a random sample from among a larger population.
First selects a fixed starting point in the larger population and then obtains subsequent observations by using a constant interval between samples taken
Hence, if the total population was 1,000, a random systematic sampling of 100 data points within that population would involve observing every 10th data point
Good for: larger populations
Can still be considered random as long as starting point was decided beforehand and was random
What is opportunity sampling?
Non probability
Using individuals who are available to us at the time
Most common method in healthcare research
Good for: convenience, cost, attrition, may be necessary in clinic
Bad for: representation and bias as volunteers may be certain kind of person. Low external validity
What is quota sampling?
Non probability
Have predefined fixed quota then select people as they come if they fit the quota
Good for: recognising the need for representation, makes sure enough people of each desired strata are recruited
Bad for: NOT random
What is a type I error?
Hallucination.
Falsely reject null hypothesis.
This is usually due to the study testing too many outcomes
We also have to choose a probability cut off but p<0.05 is saying that the likelihood of a relationship being due to chance is 1 in 20 which is well within realms of possibility
What is a type II error?
Blindness
Falsely accepting the null hypothesis
Commonly occurs when sample is not adequate. Need enough people to see effects
Too small: may not see anything
Too large: may waste resources (unethical?)
