Variance:
- Represents the amount of variability in the data
- Three main methods of representing variability o Range (smallest and largest data points) o Standard deviation o Variance
Correlation:
- Standardised representation of the associations between two variables
- Can range from -1.0 to 1.0
- Coefficient of determination is the correlation squared
Composite variables:
- Test scores tend to be based on the sum of two or more items
- g. the Beck Depression inventory consists of 21 items:
o Items are scored on an ordinal scale from 0-3 o Thus the range of scores possible is 0-63
- Such ‘sum scores’ are known as ‘composite scores’.
Composite variables and variance:
- The variance of a composite score is a function of:
o The variance associated with the individual items o The correlation amongst the items
- A positive correlation between two variables increases the variance associated with the composite derived from those two variables.
o i.e. as correlation increases (positive), the magnitude of the corresponding composite score variance also increases.
Binary items- Variance:
- Most frequently encountered in achievement type tests, e.g. dichotomous items such as exam or intelligent test items
- All other things equal, having more variance is typically better than having less variance with respect to some important psychometric characteristics
- Calculating the variance of a composite based on dichotomous items works the same way as demonstrated for ordinal/continuous items
Interpreting test scores:
Most scores produced by psychological tests are not clearly interpretable in their own right
- Two of the most common psychological test score interpretations are (1) relative and (2) abstract
- Relative interpretations are based on the analysis of data
- Abstract interpretations are based on theoretically relevant characteristics of the body of research which supports the test scores as valid indicators of a psychological construct
Relative interpretations:
- From the relative perspective, to interpret an individual’s score we need to:
o Make reference to an entire distribution of scores on the test and o Identify where the individual falls in the distribution
- In practical terms this involves knowing the mean distribution of scores and the standard deviation of the distribution of scores
Raw scores- limitations:
- Although knowledge of the mean and standard deviation associated with a group of scores allows us to interpret a particular score in a relative way, it is not a precise method unless you are a human calculator.
- Instead raw scores can be converted into standardised scores which incorporate information about the mean and standard deviation- z-scores.
Z-scores:
- Have a mean of zero and a SD of one
- To convert raw scores into z-scores: z = X – μ / σ
- By framing the meaning of a score in terms of ‘distance from the mean’, the z score frees us from worrying about the units of the original test score
- Can be used to compare scores across tests that are on different sized units
Converted standard scores:
- Ideally don’t want to use negative numbers (in z scores), so you can easily convert them into alternated standard scores
- Accomplished by rescaling the scores so that the converted scores have a different mean and standard deviation- T scores.
- With t-scores the mean is always 50 and the standard deviation is always 10
T-scores:
- Convert raw scores to z scores
Convert z scores using the formula: T=z(10) +50