Lecture 4- Study Design Flashcards
What are the two main types of study design?
- Descriptive which describe e.g. surveys
- Analytic which test hypotheses e.g. experimental or observational
What do we almost always use in descriptive studies?
A sample (not the entire population)
What is the difference between a well-defined population and a not well-defied population?
Well-defined means there is no ambiguity as to what population you want to collect from. Think about things like timings, where, what specifically etc.
What is an exception to a descriptive study always using samples?
A census (nationwide every year), or if a population is super small. The reason we use samples is simply because it is impractical to collect data from everyone (cost too much time and money)
How do we achieve a representative sample?
Random sampling. Everyone has an equal chance of selection. If sample was chosen by the individuals themselves it could mean it is not representative as certain individuals may be more likely to volunteer.
What is a sampling frame?
List of items in a population from which a sample is taken
Does a sampling frame usually coincide with an entire population?
No. Even measures like the electrical role and telephone numbers don’t encompass the far reaches of society (people not enrolled/ that don’t have numbers). And there are also cases where it is just not possible depending on what you are trying to investigate e.g. you cannot reach everyone with depression
In the case where a sampling frame can’t include everyone in your desired group what might you have to do?
Use a measure that is pretty close (e.g. electoral role to stand in for population) as proxy. You then have to think through what the consequences might be of not capturing everyone in your sampling frame and therefore your final sample.
In a simple random sample how do chances of selection differ between people?
They don’t everyone has equal chance of selection
What are the two types of error that can arise from sampling (account for differences in the true population mean and sample mean)? How is this effected by sample size?
- Random error (chance): this is due to natural variability. Increasing sample size will decrease random fluctuations in the sample mean.
- Systematic error (bias): occurs due to aspects of the study design systematically distorting the studies results e.g. there is a selection bias which stops the sample from being representative or there is information bias where individuals in the sample provide incorrect inform (lie or can’t remember). This type of error cannot be reduced by increasing sample size.
What is probability sampling?
Where we take in to account the probability of any one person being selected to be in the sample.
What is the most common probability sample?
Simple random sample.
Each subgroup of the population has an equal chance of selection and thus each individual does as well.
What is stratified sampling?
Useful when a population contains several groups of similar individuals.
For this you would take a simple random sample from within each group/ sample (instead of across the whole board)
What are the two types of stratified sampling?
- You can sample by selecting individuals from stratum but taking into account group size. In other words probability is proportional to size and everyone still has an equal chance of selection.
- Alternatively you can sample with equal numbers from each strata. So that those in smaller strata are more likely to be selected (pulling out the minorities)
What is cluster sampling?
Taking a simple random sample of groups e.g. households
