Statistical tests

What is the P value in psychology

0.05

Why would the P value be set at 0.01 in experiments

when peoples lives are at risk or when doing scientific research like using drugs

have to be sure the results weren’t down to chance- more strict

If the p value was at 0.05, what percentage of the results were down to chance

5%

What is the null hypothesis

stating there will be no difference or relationship between variables

3 reasons why you would use the sign test

• looking for a difference- the IV is directly causing the DV

• repeated measures design

• using nominal data

How to do the sign test

1- subtract or add the values from the 2 columns presented

2- add up the number of + and -

3- which ever sign has the least amount is now the S(calculated) value. e.g. number of + was 15 so s=15

4- line up N (number of participants) and the P value (0.05) and if the test is 1 or 2 tailed and go down the table to find the critical value. 

5- calculated value S must be equal or less than the critical value to be significant.

What happens in the sign test if the calculated value of S is equal or less than the critical value at 0.05

• the results are significant so we accept the alternative hypothesis and reject the null hypothesis

• occurred due to the IV causing the DV

What happens in the sign test if the calculated value of S is not equal or less than the critical value of 0.05

• the results aren’t significant so we reject the alternative and accept the null

• it is down to chance

what is nominal data

• data categorised into groups e.g. hair colour, geneder

what is ordinal data

• data is ordered, ranked on a scale and is subjective e.g. rating happiness levels on a scale

what is interval data

• rating with a scale with measurements where each unit is the same size, there is no zero and is objective- its facts e.g. your weight, temperature

9 point rule of deciding is results are significant

1. Is the Alternative Hypothesis Directional (one tailed) or Non Directional (two tailed)

2. State the Null Hypothesis and don’t forget the phrase “it is down to chance”.

3. State the probability value p ≤ 0.05

4. What is the N value (number of participants) or the df value (degrees of freedom)

5. Work out the critical value from the table and state it.

6. State the observed/calculated value given in the question or calculated by you in the Sign Test.

7. State whether the results are significant or not and how you determined this i.e. the observed/calculated value is more than or equal to or less than or equal to the critical value.

8. If significant reject the Null Hypothesis and accept the Alternative Hypothesis or if not significant accept the Null hypothesis and reject the Alternative Hypothesis.

Then in the context of the scenario

3 reasons why you would use Mann Whitney test

• looking for a difference between 2 groups

• independent group designs

• ordinal data

features of parametric tests

• detect significance in some situations where non-parametric tests can’t

• more powerful than non parametric tests as calculations use the mean and standard deviation

2 types of parametric tests

• unrelated and related t tests

when would you use unrelated t test

• test of difference

• when data is interval and independent measures

when would you use related t test

• test of difference

• when data is interval and repeated or matches pairs

info needed for unrelated t-test

• df= Na + Nb- 2

• Hypothesis (directional or non directional)

• P- value

• Observed (calculated value)

• Critical value

• Observed value > or equal to critical value

• Are results significant or not

• Conclusion- which hypothesis are you accepting

info needed for related t-test

df=N-1

Hypothesis (directional or non-directional)

P-value

Observed (calculated) value

Critical value

Observed value > or equal to critical value

Conclusion- which hypothesis are you accepting

3 reasons why you would use a Wilcoxon test

• testing a difference 

• repeated measures design or matched pairs

• ordinal data

What are the 3 correlation tests

• Chi-squared

• Spearman’s rho

• Pearson’s r

3 reasons why you would use Spearman’s rho

• tests an association between 2 variables

• uses ordinal and interval data

2 reasons why you would use Pearson’s r

• Parametric correlation

• interval data

Info needed for Spearman’s rho

N

Hypothesis (directional or non-directional)

P-value

Observed (calculated) value

Critical value

Observed value > or equal to critical value (yes or no)

Conclusion- which hypothesis are you accepting

Info needed for Pearson’s r

Df (N-2)

Hypothesis (directional or non-directional)

P-value

Observed (calculated) value

Critical value

Observed value > or equal to critical value

Conclusion- which hypothesis are you accepting

3 reasons when you use chi-squared test

• testing a difference/association between 2 variables

• nominal data recorded as a frequency count of the categories

• independent design

Info needed for Chi squared

Df (rows-1)x (columns-1)

Hypothesis (directional or non directional)

P-value

Calculated value of Chi

Critical value

Calculated value > or equal to critical value

Conclusion- which hypothesis are you accepting

Features of chi squared (contingency table)

• data is in form of a frequency count and entered in a contingency table

• contingency table involves 4 cells (2×2) but can be more.

• In a 2×2 contingency table, the 1st number means there’s 2 rows and the 2nd means there’s 2 columns

What is a type 1 error

• we assume the findings show something when they don’t.

• we reject the null which is actually true and wrongly accept alternative hypothesis

• assume results down to IV but were due to chance

When does a type 1 error occur

• when p-value is set too leniently e.g. 0.10

• because higher p-value increases possibility results down to chance

What is a type 2 error

• we miss something that’s actually happening

• falsely accept the null and wrongly reject the alternative

• assume results down to chance but were due to IV

When does a type 2 error occur 

• when p-value is too strict e.g. 0.01 

How to asses internal reliability

Use spilt half method

• compare 2 halves of questionnaire, test.

• done by randomly selecting half the test items and put them on Form A and other in Form B

• so you’ll end up with 2 form of the same test

• each form should give same score if items on test were consistent

How to assess external reliability

Test- retest

• give same test to same person on 2 different occasions

• if there’s no treatment in between tests, it should give same data

• if results aren’t similar- low reliability 

• time between tests must be long enough so person cant remember their past answers but not too long where their thoughts may change. 

what 4 factors affect internal validity

participant effect

investigator effects

situational variables

participant variables

participant effects affecting validity and how to overcome them

e.g. demand characteristics, social desirability

use single blind, lie scales

investigator effects affecting validity and how to overcome them

e.g. leading questions

use double blind, standardised instructions

situational variables affecting validity and how to overcome them

e.g. noise, temperature, time of day

use standardised procedure

participant variables affecting validity and how to overcome them

e.g. age, intelligence, experiences

use random allocation, matched pairs