Null hypothesis significance testing

Question 1

Inferential statistics

Answer 1

Attempting to see if we can:
- Infer an alternative explanation that the difference observed is NOT due to chance e.g. due to IV
-Or if it is a result of random variance caused by sampling error, e.g. the IV has no effect

Question 2

Null hypothesis symbol

Answer 2

Hₒ

Question 3

Null hypothesis meaning

Answer 3

In the population we’re sampling, there is no relationship between the variables being tested (IV effect on DV)

Question 4

What does it mean to assume null hypothesis is true

Answer 4

We say…
Under Hₒ there will be no statistically significant differences between groups (experimental condition)

Question 5

When can we reject the null hypothesis?

Answer 5

If we were to assume this is true…
But turns out the probability it is due to chance/ sampling error is actually very low
Then we can reject it

Question 6

Symbol for the alternative hypothesis

Answer 6

H₁

Question 7

What is the alternative hypothesis?

Answer 7

In the population sampled, there is a significant difference between experimental conditions scores, i.e. the IV had an effect on the DV

Question 8

If the null hypothesis is rejected, is the alternative hypothesis therefore true?

Answer 8

Maybe but maybe not
Because the tests tell us the probability of the data being obtained due to chance, assuming the null hypothesis is true
Does not tell us the probability of either hypothesis being true as a result of what data we obtained

Question 9

How can we argue for the alternative hypothesis being true?

Answer 9

Conditions are identical and controlled in every way - all confounding variables controlled for

Question 10

Why might we not be able to conclude the alternative hypothesis is true?

Answer 10

Assumption we have controlled everything is wrong
Experimenter bias, small errors e.g. we are not aware of
there may be other plausaible mechanisms and explanations for the results

Question 11

NHST

Answer 11

Null hypothesis significance testing

Question 12

What are NHSTs?

Answer 12

Allow us to test for statistically significant differences between groups = how probable this observation is if its purely due to chance
And rule out sampling error

Question 13

What do NHSTs use?

Answer 13

Z scores and p values

Question 14

Using z scores recap

Answer 14

Transform scores into z scores = how many SDs this score is away from the mean
Score taken away from mean
Divide by the standard deviation
Use a table to obtain probability of obtaining a data score with this z value/ the percentage of scores below this, above this etc

Question 15

p value

Answer 15

A probability value of obtaining any given score
This is the value looked up in tables of z value that shows probability of a score being in smaller/ larger portion/ mean to z portion

Question 16

How to use z values to determine significance

Answer 16

Based on standardised distribution…
z value of +-1.96 will be 5% of the population in total
If, given the sample size, we can find the number of Ps we will expect to have this z score (5% of the population)
If we were to randomly select participants from this sample

Question 17

How to use z values to determine p value

Answer 17

Use same principle as determining outliers:
Think of 2 conditions as having their own standardised distributions that overlap

Question 18

alpha sign

Answer 18

Threshold of significance

Question 19

NHST on paired distributions

Answer 19

How many scores would we expect above the threshold (alpha) if null hypothesis is true
E.g. what percentage of scores would need to be present in the overlap of both distributions for each condition
(If null hypothesis is true then the distributions should be very similar and have a lot of overlap)

Question 20

p value for null hypothesis significance test

Answer 20

We obtain the probability that, if assuming the H0 is true, we expect to observe results as extremely different as this a percentage of times equal to the p value
E.g if p = 0.001 then 0.1% of the time

Question 21

If the null hypothesis is true, then group differences that are extreme are…

Answer 21

unlikely but not impossible

Question 22

Equation to show means are identical aka nul hypothesis is true

Answer 22

μ₁ - μ₂ = 0
aka the means of both sample are the same

Question 23

Sampling distrubtion of a null hypothesis (plotting means of study that had been done again and again)

Answer 23

For each condition, the means make up its own singular curve:
But they are identical because all the means are the same
And as we take more samples, converges on the population mean closer and closer

Question 24

What happens if we assume the null hypothesis is true but we obtain large difference in means?

Answer 24

If we obtain a difference in means that is quite large, when plotting a sampling mean assuming the null hypothesis is true,
It is still possible to have a large difference in means even if the results are due to chance by sampling error alone

Question 25

Central limit theorem undepins NHST

Answer 25

Allows us to imagine all possible outcomes and compare this to the data we collected in single study

Question 26

Logic of NHST

Answer 26

Obtain p value = assuming the null hypothesis is true, we expect the results at least as extreme as this p% times aka due to chance is this many times If this is very small = improbable that this is sampling error Based on our subjective judgement of how small this value is, we can

Question 27

alpha

Answer 27

Threshold for statistical significance set

Question 28

Most common a valye

Answer 28

.05 = p

Question 29

If p value is less than .05 that was set as alpha

Answer 29

Then we reject the null hypothesis: threshold the likelihood of results being due to chance was exceded negatively (lower than alpha)

Question 30

p value translation to z value

Answer 30

On a z curve, our decided p value will transform to a z value which will be a certain number of SDs away from mean and also, given the area between the z values on wither side, will be a percentage This percentage = chance that we reject the null hypothesis even though it's true

Question 31

NHST process

Answer 31

State hypothesis + null hypothesis Set alpha level Apply NHST (statistical model) Decide whether to reject null hypothesis or not

Question 32

Stating hypothesis types

Answer 32

Directional Non-directional

Question 33

Directional hypothesis

Answer 33

Predicting the direction you expect group differences to occur such as smaller or bigger between group scores

Question 34

Non-directional hypothesis

Answer 34

Predict there will be significant group differences but will not state the direction this will occur in

Question 35

Types of tests

Answer 35

One tailed Two tailed

Question 36

One tailed hypothesis test

Question 37

Two tailed hypothesis test

Answer 37

Reject the null hypothesis id sample results fall in EITHER tail of the sampling distribution e.g., extreme high or lower end Alpha value set at will be split between 2 tails e.g. .025

Question 38

Tails

Answer 38

The extreme ends of the bell curve results may fall under

Question 39

One-tailed hypothesis test

Answer 39

Reject the null hypothesis if a sample falls in the PREDICTED TAIL e.g. extreme end of the sampling distribution Alpha value set at will NOT be split between 2 tails e.g. stays .05 and is less stringent

Question 40

Two tailed test is used for what?

Answer 40

Non-directional hypothesis: if we observe statistically extreme difference in either direction, then we reject H0

Question 41

One tailed test is used for what?

Answer 41

Directional hypothesis

Question 42

How to decide what hypothesis to use>

Answer 42

Directional hypothesis is chosen if we have strong empirical or theoretical basis to it from previous literature

Question 43

Pre-registration method

Answer 43

Decide what to do before the study is done: Hypothesis direction test will use Alpha level And justification for why based on pre-reading