Lect-6-Sampling and sample size

Census vs. Sample Surveys
- Census: Investigation of all animals in a population; expensive and sometimes impractical.
- Sample Survey: Investigates a subset of the population; aims to estimate variables accurately and without bias.

Target Population: Total population where information is required, ideally the population at risk.
Study Population: The population from which a sample is drawn; consists of elementary units (e.g., individual animals).
Stratum: A collection of elementary units grouped by a common characteristic (e.g., dairy cows on a farm).
Sampling Frame: A list identifying members of the study population (e.g., lists of veterinary practices).
Sampling Unit: Each member of the sampling frame.
Sampling Fraction: Ratio of sample size to study population size (e.g., 10 out of 1000 = 1%).

Convenience Sampling: Selection of easily accessible units; may lead to biased results.
- Example: Selecting first 10 cows entering a milking parlour may underestimate lameness prevalence.
Purposive Selection: Choosing units based on the desire to balance characteristics with the target population.
- Example: Selecting blood samples that represent average characteristics in a tuberculosis test; risks bias towards the mean.

Simple Random Sampling: All units in the study population are listed, and units are selected randomly.
Systematic Sampling: Units are chosen at equal intervals from a randomized starting point.
- Example: Selecting one animal from every 100 based on a random choice of the first unit.
Stratified Sampling: Divides the population into strata and samples randomly from each stratum.
- Example: Different regions in Bhutan for livestock movement sampling; proportions reflect the sample size according to the total in each region.

Cluster Sampling: Defined by geographical or other categories; involves sampling a few clusters instead of all, acceptable for managing costs and time.
- One-stage Cluster Sampling: Sampling all units within selected clusters.
- Two-stage Cluster Sampling: Selecting clusters and then sub-sampling within them.
- Multistage Cluster Sampling: Further progressing to additional sampling levels (i.e., regions, then farms, then individual animals).

Non-statistical Factors: Availability of manpower and sampling frames.
Statistical Factors: Desired precision of the prevalence estimate and expected disease prevalence.

Absolute Error: Acceptable range of prevalence (e.g., 40% with a 2% error ranges from 38% to 42%).
Relative Error: Error expressed proportionately; a 2% error of a 40% prevalence results in a slight variation (39.2%-40.8%).

Prior estimates of prevalence can guide sample size determination; close approximations to 0% or 100% require fewer samples compared to a prevalence around 50%.