Definitions of terms ✅

Question 1

What are the scales of measurement?

Answer 1

Categorical - nominal (e.g. gender):
- Contains labels
- No numerical relationship between values

Discreet or Continuous:

1. Ordinal (e.g likert scale):
   - Numerical relationship
   - Data organised by rank
   - Equal/unequal intervals between values
2. Interval (e.g. shoe size)
   - Equal intervals between values
   - Numerical relationship
   - Scale has no true zero point
3. Ratio (e.g. distance)
   - Numerical relationship
   - Equal intervals
   - True zero point

Question 2

What are the descriptives statistics and when is each type of statistic used?

Answer 2

Definition: summarise a set of sample values

Two forms of statistics:
1. central tendency
2. spread

When to use:
1. Discrete/continuous data AND normally distributed -> Mean & SD
2. Discrete/continuous data NOT normally distributed -> Median & Range
3. Categorical data: Mode

Question 3

Which experimental design can infer causality?

Answer 3

True experimental IVs:
- IVs actively manipulated
- Random allocation possible
=> can make claims about causality

Quasi-experimental IVs:
- IVs reflect fixed characteristics (e.g. smoker/non-smoker)
- Random allocation not possible
=> Cannot be sure about implying causality

Question 4

Which statistics test to use when:

Answer 4

Experimental design
1. One IV only:

• Only 2 levels:
  -> independent vs. paired t-test (parametric)
  -> Mann-Whitney U test vs. Wilcoxon test (non-parametric)
• More than 2 levels:
  -> 1-way independent vs. repeated measure ANOVA (parametric)
  -> Kruskal Wallis vs. Friedman test (non-parametric)

2. At least 2 IVs:
   * 2-way independent/ repeated measure/mixed ANOVA
   -> no non-parametric equivalent

Correlation/relationship:
All types uses Pearson’s r for correlations

• Normal correlation (non-parametric):
  -> Spearman’s rho (N>20)
  -> Kendall’s Tau (N<20)
• Partial correlation
• Regression
  => The remaining two have no non-parametric equivalent

Categorical data:

• One-variable Chi-Square
• Chi-Square test of independence (e.g. 2*2)

Question 5

What are the features of a normally-distributed population?

Answer 5

• Features: bell-shaped AND symmetrical about the mean
• Types of kurtosis:
  1. Mesokurtic: balance
  2. Platykurtic: large s.d.
  3. Leptokurtic: small s.d.
• Types of skew:
  1. No skew: perfect central
  2. Positive skew: curve leans to left -> more +ve extremes
  3. Negative skew: curve leans to right -> more -ve extremes
• Non-normal distribution:
  1. Bimodal: 2 peaks
  2. Uniform: no peaks
  -> Use non-parametric tests

Question 6

What can be said about sample statistics?

Answer 6

• Sample statistics can be used to infer population parameters
• Sampling error (SE): degree to which sample statistics differ from population parameters
  -> Estimated Sampling error (E.S.E): since we don’t know population distributions
• Minimalising errors - sample must be:
  1. Representative (randomly selected)
  2. Sufficient in size

Question 7

What is confidence interval?

Answer 7

• Confidence Intervals (CIs) are interval estimates of
  population parameters
• 95% of all sampled means will fall WITHIN the 95% bounds of the population mean (μ ± 1.96 SE)

Question 8

What are Type I and Type II errors?

Answer 8

1. Type I error: reject null when the opposite is true
2. Type II: fail to reject null when the opposite is true
   -> setting α threshold too low (e.g. α = 0.01)

Question 9

What are the assumptions of a normally distributed population?

Answer 9

1. Normality: the DV should be normally distributed, under each level of the IV
2. Homogeneity of variance: the variance in the DV, under
   each level of the IV, should be equivalent -> Levene’s test (if significant then VIOLATED)
3. Equivalent sample size
4. Independence of observations: Scores under each level of the IV should be independent

=> 2 and 4 not relevant for repeated measure designs
=> If violated any SHOULD use non-parametric equivalents