2. Sampling and Sample Size

The animals we use in a study

What does the study sample consist of?

- Non-random sampling (non-probability)

- Consecutive sampling (non-probability)

- Census sampling (non-probability)

- Random sampling (probability)

- X

What are some of the ways of getting a good sample from the "available" population? (5)

- Convenience (go out and get animals who are easy and convenient to find)

- Volunteer (owners volunteer their animals for your study; may not represent the population in general)

- Purposeful (look for cases and purposefully select the animals that you want)

- Judgement (arbitrary idea of who you will and won't accept)

What are the 4 types of non-random sampling?

consecutive

What type of sampling is this: Deciding to study the next X number of cases that you see

Case series

What type of study is consecutive sampling often used for?

census

What type of sampling is Including all cases in a finite population OR populations with known data

- Small populations (ex: endangered species)

- "Big data"

When is census sampling used?

random

What type of sampling is All animals selected have an equal chance of being chosen.

F (humans always have some form of bias)

T/F: We as humans are able to perform random sampling

F ( — you still need to check and evaluate.)

T/F: Random selection guarantees identical study groups.

simple

What type of random sampling is " identified animals are selected from the available population using a random number table or software that generates random numbers

Systematic

What type of random sampling is defined as - selecting from a randomly selected first subject, then sample every x number case

cyclical or seasonal patterns

What conditions would systematic random sampling not work

Stratified

What type of random sampling is defined as picking groups (like male/female, age etC) and their random selection within each grouping.

clustered

What type of random sampling is defined as using natural groupings (herds, owners, zoo exhibits, locations) and then assigning different groups randomly but assign all the animals in the group to the same treatment course

-Randomization

- Matching (controls are individually selected to match cases)

- Restriction (available study population is artificially limited to particular characteristics)

What are the 3 sampling methods used to control bias?

Case-control studies

When is matching most commonly used as a sampling method to help control bias?

- Start with healthy animals

- Randomly select animals to receive intervention/treatment or not

- Look at the outcome

Describe how to sample for a randomized clinical trial.

- Start today with healthy animals

- Separate naturally based on exposure status (you are NOT assigning this; this happens organically)

- Look prospectively at the outcome

Describe how to sample for a cohort study.

- Start with diseased animals (outcome) today

- Choose a group of very similar, but healthy animals

- Look back in time retrospectively at their exposure status and compare the diseased animals to a similar group of healthy animals

Describe how to sample for a case-control study.

- Determine a total number of animals needed to sample

- Survey a group of animals today

- Look at their disease status AND exposures at the same time

Describe how to sample for a cross sectional study.

based on disease status.

How are study samples selected in a case-control study?

based on exposure to a potential risk factor

How are study samples selected in a cohort study?

- Underpowered (—> non-significant findings

- "Our results were almost significant; if we had a larger study, we might have found significance." )

What are the problems with a very small study?

Only a 2x increased risk

If you have a very large study, what does an increased RR or OR of 100% mean?

That the risk was 2 or more times larger.

With very large studies, what will more impressive risk factors say?

- Forces specification of outcomes

- Leads to a stated recruitment goal

- Encourages development of timetables and budgets

- Discourages small, inconclusive studies

- Helps prevent poor animal welfare issues

What is the importance of sample size? (5)

- No discussion of sample size

- Unrealistic assumptions and estimates

- Failure to explore sample size for a range of values

- Failure to account for attrition

What are common mistakes regarding sample size? (4)

- When studies try to add or subtract samples in order to get a "good" result

What is "p-hacking"?

T ( Don't do "p-hacking"; it is bad)

T/F: Don't change sample classification or remove outliers without excellent reasons and documentation.

- Confidence and power

- Estimated "available population" size

- Planned ratio of controls/cases

- Estimated positive controls (aka expected prevalence of animals not exposed)

- Estimated positive cases (aka expected prevalence of animals exposed)

- Estimated error or precision (of your instrumentation)

What information may you need for sample size calculations? (6)

- Confidence = 95%

- Power = 80%

What is the usual confidence and power for sample size calculations?

False — different studies need different information to calculate sample size.

T/F: All types of studies need the same information in order to calculate sample size.

- Protects against lost animals or samples (gives you a little wiggle room)

- Add 25%

After calculating the needed sample size, adding a few extra samples is customary? Why? How much is usually added?