Correlation, observation and self-report analysis

Correlation

A correlation measures the strength and direction of a relationship between two co-variables, from $-1$ through $0$ to $+1$. Positive means both rise together; negative means one rises as the other falls; near zero means no linear relationship. Plot it as a scattergram.

The rule everyone quotes and few apply

Correlation does not establish cause. Three reasons, and an answer that gives all three earns full marks: the causal direction may run the other way; a third variable may cause both; and the relationship may be coincidence, especially with a small sample. Say which third variable might be operating in the study in front of you — that is what separates a memorised line from an analysis.

Observation data

Behavioural categories must be defined so precisely that two observers code the same event the same way; vague categories destroy reliability. Event sampling counts every occurrence of a behaviour; time sampling records what is happening at fixed intervals — cheaper, but it misses behaviour between the intervals. Inter-rater reliability is checked by correlating two observers' records.

Self-report data

Closed questions give quantitative data that is quick to analyse but forces answers into fixed boxes; open questions give qualitative depth but are slow and subjective to analyse. Both suffer social desirability bias, and leading questions damage validity.

Analysing qualitative data

Thematic analysis codes the text and groups codes into themes, keeping meaning that a number would lose. Its weakness is subjectivity, reduced by having a second researcher code independently and by reporting the coding rules.