Stats G1.5 Final Part 1 (new tests)

Null Hypothesis Statistical Testing (NHST)

How likely are these results to have occurred by chance? Look at p-values

Sampling distribution

The disribution of all possible values of that statistic that would obtained if an infinite number of samples of the same size were drawn from the population described by the null.

Family-wise error rate

• multiple tests= increased chance of type 1 error

• studies with multiple outcomes are more prone to false positives

Bonferroni adjustment

• Correction used to reduce familyiwise errors

• divides alpha by number of tests

• increases risk of type II erorr (false null)

Effect size

Tells us magnitude of an effect

When to use One-way Repeated Measures ANOVA?

• IV

  • Nominal/Ordinal

  • Within-Subjects

  • 3 or more levels

• DV

  • Measured on I/R scale

Between-subjects ANOVA vs. One-way repeated measures ANOVA

• Repeated Measures ANOVA

  • we assume that the datapoints are NOT independent of one another (assume there is a correlation)

  • Use F statistic that compares between “group” variance to within-subject variance

Hypotheses for One-way repeated ANOVA and Between-subjects ANOVA

H₀ : x̄₁ = x̄₂ = x̄₃…

H₁ : x̄₁ ≠ x̄₂ ≠ x̄₃…

Statistical power

• The likelihood you have of successfully detecting a difference (or relationship) that actually exists in your data

• Typically you want to have at least 80% power

Ways to conduct power analysis

• Cohen ‘92 article

• G* Power

What is a factorial ANOVA?

Compares the mean differences between groups that have been split on two or more IVs (called factors)

Primary purpose of factorial ANOVA

understand if there is an interation between the 2+ IVs on the DV

When to use two-way Between-Subjects ANOVA

• Two nominal between-subjects

• IVs with 2 or more levels

• DV that is measured on an I/R scale

Effects in two-way b/s ANOVA

• Main effects (one for each IV)

• Interaction (between the IVs)

What is Multiple Linear Regression?

• Uses two more more IVs to predict the outcome of a DV

When to use multiple linear regression?

• To determine relationship between

  • One DV that is IR AND

  • Two ore more IVs that are IR or dicotomous

Goals of multiple regression

• Prediction

  • EX: insurance companies predict who will crash based on age, sex, zip code, etc.

• Explanation

  • attempting to understand a phenomenon by examining its relationship with a group of variables

    • EX: finding which variables are most strongly associated with binge drinking