Hypothesis testing - RM2

Hypothesis testing

the process of determining whether data support your prediction

supported or not supported

Hypotheses aren't proven "true" or "false" - all about probabilities

H1: Your research hypothesis​

"there is an effect"​

aka. alternative hypothesis​

the one that you believe is "true"​

goal: retain the alternative hypothesis​

​

H0: The null hypothesis​

"there is no effect" (e.g., no difference between means)​

the one that, statistically, you are actually testing​

goal: reject the null hypothesis​

Steps in hypothesis testing

1. We determine what the population (distribution) would look like if the null hypothesis were true​

2. We see if our sample data are likely to have come from this distribution​

3. If it is unlikely that our data came from the null hypothesis distribution, we reject the idea that the null hypothesis is the best way to describe our sample​

4. Because the null is false, we accept that our hypothesis (the alternative hypothesis) is a better way to describe our data​

Practice steps

1.State the null and alternative hypotheses.​

These actually refer to predictions about the population.​

2. Determine the characteristics of the null distribution.​

Alpha (region of rejection)?​

One- or two-tailed?​

3. Calculate the appropriate test statistic.​

Common options: z, t, F, r...​

4. State the decision about whether we can reject the null hypothesis.​

Is p less than .05? OR is the test value more extreme than the critical value? If so, the null hypothesis is rejected.​

5. State the conclusion verbally in terms of the finding and variable names.​

One-tailed (or directional) test

predict direction of the effect (i.e., mean X higher than mean Y)​

Two-tailed (or non-directional) test

no prediction about the direction of the effect (i.e., means differ but not specified which higher)​

Type I error

Null is true, Reject null

False alarm/ false positive

Crying wolf

Correct decisions

Null is false and reject null

Null is true and fail to reject null

Type II error

Null is false and fail to reject null

false negative

fail to announce the wolf

Type I error rate

alpha at .05

inversely related

type 1 and type 2 error

p-value

is the probability of obtaining a test statistic (e.g., r-value, t-value) at least as extreme as the one that was actually observed, given that the null hypothesis is true.​

significant p value

want p-value less than .05