Significance Testing

Question 1

Hypotheses

Answer 1

Never predict no change.
Never right insignificant.
Results may be “not significant”.
Alternative hypothesis- to show IV having an effect.
Null hypothesis- to show IV having no affect. Not the opposite effect.

Question 2

Tailed hypotheses

Answer 2

One tailed- predicts direction.

Two tailed- predicts a change without direction.

Question 3

Available tests

Answer 3

Parametric tests for differences between two groups- between subject t-test; within subject t-test.
Parametric tests for correlations between two groups- pearson’s r.

Correlational study- no IV and DV.

Question 4

Checks for normality

Answer 4

Large sample- can plot histogram and see if data is symmetrical.
Check central tendency- mean and median approx the same.
Skewness value should be less than half standard error value.
Kurtosis score should be less than half too.

Question 5

Checks for normality. Kolmogorov-smirnov test

Answer 5

If you don’t have large sample or data doesn’t appear normal.
Compares set of scores with normally distributed set.
We don’t want our data to be significantly different.
P

Question 6

Experimental method

Answer 6

Formulate hypothesis. 
Design way of measuring prediction. 
Think about confounding factors. 
Measure DV. 
Compare statistics for two groups. 
Decide if difference is because of the IV or chance.

Question 7

Significance testing

Answer 7

Find difference between two sample means.
Confounding errors- have systematic effects. Can be controlled.
Random errors- have unsystematic effects. Cause unpredictable differences in scores.

Question 8

Accepting and rejecting hypotheses

Answer 8

Spss converts difference mean to a z score and calculates how many std devs away from the mean.
Calculates likelihood we got this size difference due to random errors.
Very likely- retain the null.
Not likely- reject the bulk; accept alternative.

Question 9

Alpha or p values

Answer 9

Cut off point is 5%.

P.05 - more likely to have happened by chance.

Question 10

Errors

Answer 10

Type 1- we accept we find a difference when it’s actually down to random errors.
Type 2- we accept the difference was due to random errors when it was actually due to the IV.

Question 11

Interpreting results

Answer 11

Accepting alternative hypothesis- accepting the IV works as predicted.
BUT
Never proves the null hypothesis false. Could always be due to an unrepresentative sample or confounding variables.