Lecture 11: Bayesian Inference Flashcards
(32 cards)
What is the goal of hypothesis testing in psychology?
To determine if observed variation is due to manipulation or random noise.
What are the two main sources of variation in behavioural data?
Systematic and random variation.
What kind of variation results from manipulated variables?
Systematic variation.
What kind of variation results from chance or unmeasured factors?
Random variation.
What does a p-value represent in NHST?
The probability of the data assuming the null hypothesis is true.
What is a common p-value threshold for rejecting the null hypothesis?
p < .05.
What logical fallacy is NHST vulnerable to when misused?
Misapplying modus tollens probabilistically.
What incorrect belief often results from a low p-value?
That the null hypothesis is false.
What probability does NHST actually provide?
P(Data | H₀).
What probability do researchers typically want?
P(H₀ | Data).
What hypothesis does NHST fail to evaluate directly?
The alternative hypothesis (H₁).
What analogy is used to show problems with NHST logic?
HIV testing in low-prevalence populations.
What statistical concept does the HIV test analogy illustrate?
The importance of base rates (prior probability).
What key issue does NHST ignore, as shown in the HIV analogy?
Base rate of the hypothesis being true.
What does Bayesian inference compute?
The probability of a hypothesis given the data.
What three components make up Bayes’ theorem?
Prior probability, likelihood, and posterior probability.
What does prior probability represent?
Beliefs about a hypothesis before seeing the data.
What does posterior probability represent?
Updated belief in the hypothesis after seeing the data.
What is the purpose of Bayesian updating?
To revise prior beliefs based on new evidence.
What example is used to explain Bayesian updating?
Coin tosses with fair and biased coins.
What do Bayes Factors compare?
Support for H₀ versus H₁.
What type of evidence do Bayes Factors provide?
Graded evidence (e.g., anecdotal, moderate, strong).
What do Bayes Factors not rely on?
Arbitrary cutoffs like p < .05.
What is a credible interval in Bayesian statistics?
A range of values most plausible given the data.
