PSYC2012 Variables & Measurement

Variable vs. Constant

Variable Anything that varies

> Quantity that can change: phenomenon for which an objective measure has been provided.

Constant Anything that does not vary (e.g. experimental controls).

DV vs. IV

Dependent Variable What is measured – the outcome or response variable.

NOTE: Question of appropriateness/practicality – what if pain self-efficacy or pain interference has more utility than measuring pain itself?

Independent Variable A potential cause of the dependent variable which is manipulated by the researcher.

Psychological Constructs & Operationalisation

Hypothetical/Psychological Construct

An abstract concept used to describe & explain mental processes, behaviours or traits that cannot be directly observed.

 

Operational Variable

RE validity: Defining phenomena in terms of precise procedure taken to measure it.

Scales of Measurement

1. Nominal

2. Ordinal

3. Interval (& Quasi-Interval)

4. Ratio

Categorical vs. Continuous Scales

Categorical (Qualitative difference)

> Nominal & Ordinal

Continuous

> Quantitative info which can take any value in a range or interval with direction

> Measurement vs. counts

> Interval & Ratio

Nominal Data

Data can be categorised but not measured, ranked or ordered.

> "Unordered labels"

> "Same vs. different"

> E.G. Gender, hair colour, ethnicity, country, marital status

> Mutually exclusive

> Also includes binary data (yes/no, employed/unemployed)

Ordinal Data

Data categories can be ordered but differences cannot be determined (or are meaningless)

> Uneven intervals (difference between 1 & 2 not the same as 2 & 3)

> Qualitative aspect: assigning numbers but value is not absolute

> Ordinal data can be scored & treated as interval

> E.G. Rank position (1st, 2nd, 3rd)

> E.G. Tall vs. short

> E.G. Ratings or Likert Scales

Interval Data

Ordered numerical values with each unit representing equal change but lacks a true or absolute zero (so not equal ratio).

> E.G. time

> E.G. temperature

> E.G. IQ

>E.G. decibels

Arbitrary Zero: Does not mean absence of variable.

Ratio Data

Data with an absolute 0. Ratios are meaningful. E.G. Length, Width, Mass, Distance, Time

Validity

Broad Definition The degree to which a claim is true or correct

 Psychometric Definition The appropriateness, usefulness of meaningfulness of test (measurement) scores & their interpretations.

(1) The attribute must exist: reification.

(2) Variations in the attribute causally produce variation in the measurement outcomes.

Reliability

> Stability of measurement outputs across time or context.

> How well does a measurement assess an attribute?

> The degree of absence of fluctuations that are unaccounted for by noise/random error. 

Types of Reliabiltiy

1. Internal consistency

2. Inter-Rater

3. Test-Retest

Test-Retest Reliability

Stability of scores over time

> Requires testing at multiple time-points

> Affected by:

A. Dropouts/non-response rates

B. Temporal instability of constructs

C. Optimal time interval (interval should be short enough to not show true change, but not too short that Ps remember answer)

Internal Consistency

Degree of consistency among test items - that they measure the same underlying construct.

> Requires multiple items collected on one occasion

> Affected by no. of items, unidimensionality of scale, correlation b/w items

Face Validity

Face value! Extent that test appears to measure what it is intended to.

> NOTE: a bit of deception may be good to prevent "faking good."

> NOTE: Not statistical!

Construct Validity

Degree to which constructs possess a sound/coherent theoretical foundation which is operationalised through measurable descriptors

> Convergent

> Divergent

Content Validity

Tests adequately cover the content area they claim to represent = whole/unbiased domain representation (e.g. make up of final exam).

> NOTE: Not statistical!

Convergent Validity

High correlation between:

1. Items which make up same or related constructs

2. Tests that measure the same or related constructs

Divergent (Discriminant) Validity

Low correlation between:

1. Items that make up unrelated constructs

2. Tests that assess unrelated constructs

Criterion Validity

The extent to which a measure correlated with outcome criteria:

1. Predictive: Criterion in future: E.G. aptitude test vs. future job performance.

2. Concurrent: Criterion in present. Agreement with another validated assessment. E.G. depression score vs. current diagnosis