BIOL309

Question 1

Replication v Pseudoreplication

Answer 1

Replication is the process of repeating experiments to ensure accuracy.

Pseudo-replication multiple measurements are taken from non-independent samples.

Question 2

Randomisation and interpersion

Answer 2

R - Assignment treatments to experimental units to avoid bias.

I - Distributing treatments evenly across experimental units to avoid confounding variables

Question 3

Accounting for Sources of Variation/Error Objective

Answer 3

To identify and control sources of variation to improve the accuracy and reliability of experimental units

Question 4

Accounting for Sources of Variation/Error Methods

Answer 4

Blocking, randomisation, and covariates in analysis

Question 5

Assumptions of Parametric Tests

Answer 5

Normality: Data should be normally distributed.
Homogeneity of Variance: Variances should be equal across groups being compared.
Linearity and Independence: Relationships should be linear, and observations should be independent.

Question 6

Regression and ANOVA as Related Linear Models

Answer 6

Regression: Analyzes relationships between variables by fitting a line through data points.

ANOVA (Analysis of Variance): Compares means among groups to see if at least one differs significantly.

Question 7

Multiple Regression and Multifactor ANOVA

Answer 7

Multiple Regression: Extends simple regression to include multiple predictors.

Multifactor ANOVA: Analyzes the effect of two or more factors on a response variable.

Question 8

General Linear Models Combining Categorical and Continuous Predictors

Answer 8

Definition: Models that incorporate both continuous and categorical predictors to explain variance in the response variable.

Question 9

Assessing Model Fit and Assumptions Techniques

Answer 9

Use residual plots, R-squared values, and diagnostic tests to assess fit

Question 10

Assessing Model Fit and Assumptions Check

Answer 10

Ensure assumptions like normality, independence, and homoscedasticity are met

Question 11

Model Simplification and Parsimony Objective

Answer 11

Simplify model without losing explanatory power by removing non-significant predictors

Question 12

Model Simplification and Parsimony Principle

Answer 12

Prefer simpler models that adequately describe the data.

Question 13

Comparing Models Using AIC (Akaike Information Criterion) - AIC definition

Answer 13

A measure used to compare models; lower AIC indicates a better model fit with fewer parameters.

Question 14

Sample Size Considerations Importance

Answer 14

Larger sample sizes generally increase statistical power but require more resources

Question 15

Different Sampling Schemes

Answer 15

Random Sampling: Every member has an equal chance of being selected.

Stratified Sampling: Divides population into strata and samples from each stratum.

Question 16

Relationship Between Sample Size, Statistical Power, and Detectable Effect Sizes

Answer 16

Larger sample sizes increase power, allowing detection of smaller effect sizes with greater confidence.

Question 17

Completely Randomized Designs

Answer 17

Assign treatments randomly across all experimental units without restriction

Question 18

Randomised block design

Answer 18

Group experimental units into blocks before randomising treatments within each block to control for block effects

Question 19

Factorial designs

Answer 19

Study the effect of two or more factors simultaneously by varying them together in an experiment

Question 20

Nested Designs

Answer 20

Used when there are hierarchical structures in data; units within groups receive different treatments

Question 21

Repeated measures design

Answer 21

Subjects receive multiple treatments over time, allowing for within’ subject comparisons

Question 22

Rank-based Tests as Alternatives When Parametric Assumptions Are Violated

Answer 22

Examples include Mann–Whitney U test and Kruskal–Wallis test; used when data do not meet parametric assumptions like normality.

Question 23

Extending Linear Models to Non-normal Error Distributions

Answer 23

Generalized linear models (GLMs) allow for response variables with error distribution models other than normal, such as binomial or Poisson distributions.