Statistical testing

Question 1

What are inferential statistics

Answer 1

A type of statistical analysis where the formal purpose is to determine the likelihood that the effect (difference/relationship/association) found in a study is due to chance
Increases scientific credibility and objectivity of research, allowing the researcher to determine if the null hypothesis should be accepted or rejected.

Question 2

What significance level is normally used

Answer 2

p<0.05, use this unless otherwise stated

Question 3

What is a one tailed test?

Answer 3

A directional hypothesis

Question 4

What is a two tailed test?

Answer 4

A non-directional hypothesis

Question 5

When to use the sign test

Answer 5

1. When the researchers are looking for a difference between their conditions
2. A related (participants linked) design (repeated measures or matched pairs)
3. The level of measurement is nominal data, look at the dependant variable.

Question 6

How to calculate the sign test (Holiday happiness example)

Answer 6

1. Identify your 3 categories (1 = happier after, 2 = happier before, 3 = equally happy)
2. Calculate number of participants in each category, and the total participants which is N (N=14)
3. Assign category where there is no difference a 0 sign, these participants are then removed (people who are equally happy are removed), N is now 13
4. Assign categories a + or -
5. Out of the +/- categories identify which has the smallest number, this is S.

Question 7

Conclusion model

Answer 7

The calculated value of ______ is greater than/smaller than/equal to the critical value of _______ (p<________, _____ - tailed test, N =________). This means that the result is/is not significant. This means that we can accept/reject the null hypothesis that ________________________________. [If your result is significant, you then add] This means that we can accept the alternative hypothesis that _________________________________________________________. However, because the significance level was __________, there is still a _______________ probability that the results would have occurred even if _____________________.

Question 8

How to determine which statistical test is needed

Answer 8

1. What is the level of measurement
2. Is it a test of different, correlation or association?
3. What is the experimental design

Question 9

How to remember the decision table

Answer 9

Can Simon Cowell Make Winners Sing Under Real Pressure
C - Chi-squared
S - Sign test
C - Chi-squared
M - Mann-Whitney
W - Wilcoxon
S - Spearman’s rho
U - Unrelated t test
R - Related t test
P - Pearson’s r

Question 10

Order of level of measurements on the table (top to bottom)

Answer 10

Nominal, ordinal, interval

Question 11

Order of columns on the top of the table (left to right)

Answer 11

Level of measurement, Test of different (unrelated, related), Test of correlation or association

Question 12

How to work out if it’s a test of difference, correlation or association

Answer 12

Difference if for experiments, they’re looking to change the DV
Correlations are relationships where both variables are ordinal or interval
Associations are relationship where both variables are nominal

Question 13

Related or unrelated?

Answer 13

Related is when a matched pairs or repeated measures experimental design is used
Unrelated is when an independent groups design is used

Question 14

How to check for a type 1 error

Answer 14

Change the significance level to the smallest one you can (p<0.02), check if the value is still significant or not
If it is, you can be confident you haven’t made a type 1 error. If it’s not you likely have made a type 1 error

Question 15

What is a type 1 error?

Answer 15

When the null hypothesis is rejected and the alternative hypothesis is accepted when the null is ‘true’. Often when the significance level is too lenient. Likelihood of making a type 1 error is the same as the significance level

Question 16

What is a type 2 error?

Answer 16

When the null hypothesis is accepted but should have been rejected as the alternative hypothesis is ‘true’

Question 17

Type one error model

Answer 17

Currently, we are approximately ______% confident in our rejection of the null hypothesis because a ________ significance level has been used. This means that there is still a less than _____% probability of rejecting the null hypothesis when it is true and making a type 1 error. To check this, a more stringent significance level of ______ should be used which reduces the probability of a type 1 error having been made to less than ____%, making us more confident that a type 1 error hasn’t been made. When this significance level is used, the calculated value of ____ is greater than/less than/equal to the critical value of _____ (p<________, ____-tailed test, N/df = _____). This means that the result is still significant/is no longer significant. This suggests that a type 1 error has been made and so the null hypothesis should be accepted/this suggests that a type 1 error is unlikely to have been made as there is only a less than ___% probability of this and so we can be more confident in rejecting the null hypothesis.

Question 18

Extra notes about statistics

Answer 18

In Wilcoxon, if a matched pairs design is used N = the number of pairs, not number of participants
Ignore the sign when interpreting critical values tables (it only tells you the direction)

Question 19

How to draw a contingency table

Answer 19

Normal table of variables against variables, with totals at the end of the rows and columns