Research methods - statistical tests -

Question 1

What is a correlation?

Answer 1

An association between 2 continuous variables.

Positive and negative correlations.

Question 2

What is a correlation analysis?

Answer 2

Measures the link/relationship between 2 variables.

To do this we need a correlational hypothesis. (states expected association)

Question 3

How can correlations be shown?

Answer 3

Scattergram.

Question 4

How can you measure the association between 2 variables?

Answer 4

Correlation coefficient - a number between -1 (negative correlation) and +1 (positive correlation) which tells us how closely the variables are related.
Then a table of significance is used to tell us if the correlation is significant.

Question 5

What is the difference between type I and type II errors?

Answer 5

Type I error - when researcher rejects null hypothesis that is true.
Type II error - when researcher accepts null hypothesis which is false.

Question 6

What are the measures of central tendency?

Answer 6

Mean - add up all data items and divide by how many there are, only used on ratio and interval data.
Median - middle value in ordered list, used on ratio, interval and ordinal.
Mode - most common data item, used on nominal, interval and ordinal.

Question 7

Evaluation of measures of central tendency?

Answer 7

Mean: + most sensitive - easily distorted - cannot be used on nominal data.
Median: + not affected by extreme scores + easily calculated - not as sensitive.
Mode: + unaffected by extreme values + useful for discrete data - not useful when several modes.

Question 8

What are the measures of dispersion?

Answer 8

Range: distance between highest and lowest values in set of data.
Standard deviation: measures how much scores spread above and below mean. - measured in standard distribution curve.

Question 9

Evaluation of measures of dispersion?

Answer 9

Range: + easy to calculate - affected by extreme values - ignores distribution of numbers.
Standard deviation: + precise + not difficult to calculate with calculator - hides characteristics of data.

Question 10

What are the levels of measurement?

Answer 10

Different kinds of data:
Nominal - data in separate categories 
Ordinal - data ordered in some way. 
Interval - data measured using units of equal intervals. 
Ratio - true zero point.

Question 11

What are the different ways of displaying quantitative data? 5

Answer 11

Tables - measurements collected as raw data.
Bar chart - useful for nominal data.
Histogram - useful for interval data.
Line graph - useful for continuous data.
Scattergram - used for correlational analysis.

Question 12

What is the normal distribution?

Answer 12

Equally distributed around the mean on a standard distribution curve.

Question 13

What are the skewed distributions?

Answer 13

Not equally distributed around mean.
Positive skew - extreme high scores.
Negative skew - extreme low scores.

Question 14

What is a statistical test used for?

Answer 14

To determine whether a correlation found in a particular investigation is statistically significant by using a table of critical values - to find correct you need directional or non - directional hypothesis, number of participants and level of significance.
There are different tests based on conditions.
If statistical test is not significant then ‘null hypothesis’ is accepted.

Question 15

What is the ‘sign test’ and when do you use it?

Answer 15

Analyses the difference in scores between related items (same participant tested twice).
Use it when:
- looking for difference between groups of data.
- for repeated measures.
- nominal data.

Question 16

What is the level of probability in psychology?

Answer 16

0.05 (5%)

Question 17

How do you do a sign test?

Answer 17

Nominal data put in a table and work out whether participants scores were higher or lower than in second condition - if less than, negative sign, more than, positive sign.
Add up pluses and minuses and take the less frequent sign - this is called ‘S’ - and this is your calculated value.
Compare calculated value with critical value from table.
THE CALCULATED VALUE HAS TO BE EQUAL TO OR LOWER THAN CRITICAL VALUE FOR RESULT TO BE SIGNIFICANT.

Question 18

What are the different statistical tests and how do you choose which one?

Answer 18

Parametric test: PREFERRED

• unrelated t-test (INTERVAL, unrelated design)
• related t-test (INTERVAL, related design)
• Pearson’s r (INTERVAL, test of association/correlation)

Non - parametric:
- Chi-squared (NOMINAL, unrelated design, test of association/correlation)
- Sign test (NOMINAL, related design)
- Mann-Whitney (ORDINAL, unrelated design)
- Wilcoxon (ORDINAL, related design)
- Spearman’s rho. (ORDINAL, test of association/correlation)
RULE OF R - those with ‘R’ in their name, are those where the calculated value must be equal to or more than critical value to reject null hypothesis (results significant)

Question 19

What are the criteria for choosing a statistical test?

Answer 19

test of association or correlation?
Related design (repeated measures, matched pairs)?
Unrelated design (independent groups)?
Level of measurement?