STATS4- Null hypothesis significance testing

5 steps for null hypothesis significance testing

1. hypothesis

2. data

3. evaluate inconsistency with H0

4. reject or fail to reject

5. interpret the findings (in terms of the hypothesis)

2 types of hypothesis

• null (H0)

• research/alter (H1)

2 types of tailed hypothesis

• 1 tailed (directional)

• 2 tailed (non directional)

1 tailed hypothesis

• left/lower (hypothetical mean < given value)

• right/higher (hypothetical mean > given value)

2 tailed hypothesis

• hypothetical mean > given value OR hypothetical mean < given value

• 2 z scores (± the same number)

• * z score probability by 2

what is the null hypothesis

• no difference

• the sample won’t have an atypical mean from the parent population

• H0 = sample came from a population where hypothetical population mean & sample mean are equal

what is the research hypothesis

• there is a difference

• H1 sample came from a population where hypothetical population mean is higher than sample mean

what data do you need to collect from step 2

sample mean & population parameters

how to evaluate data in step 3

• assume H0 is true so sample & population mean are the same

• calculate sample z score

• large probability = not inconsistent with H0

• small probability = inconsistent with H0

when do you reject the null hypothesis

p < 0.05 (data is inconsistent with H0)

when do you fail to reject the null hypothesis

p > 0.05 (data is consistent with H0)

type 1 & type 2 errors table (draw)

what is a type 1 error

• rejecting the null hypothesis when its true

• happens naturally (5% by random sampling error)

what is a type 2 error

• failing to reject the null hypothesis when the null hypothesis isn’t true

• happens when there is a problem with the study (bias sample, sample size too small, error in experimental tasks)