3: OPERATIONALIZING VARIABLES AND RESEARCH DESIGNS

Variables

• Something that varies (must have 2+ categories/levels)

  • Can be manipulated or observed/measured

  • e.g., Anxiety: Why do some people get nervous when they speak to others?

• The way we measure a variable (type of scale used) affects the type of statistics we can use.

There are 4 different ways of measuring variables

• Nominal

• Ordinal

• Interval

• Ratio

Nominal

• Levels/categories have different names or labels and are not related to each other in a systematic way

• No quantitative difference between categories. Arbitrary values assigned to each variable level (for analysis purposes)

• Used to examine frequencies or group differences

• An experimental IV is always nominal (treatment and control categories)!

• Examples: College major, occupation, gender, religion

Ordinal Data

• Levels/categories of a variable are organized in an ordered sequenced (rank-ordered data)

• Directional differences between levels, but the magnitude of difference between levels unknown (subtraction not possible)

• Examples: class standing, SES, reading grade level, letter grades

Interval data

• Difference between the numbers of a scale are meaningful (subtractions can be made)

• Intervals between the numbers are equal in size

• No meaningful zero reference point (no ratio comparisons)

• Often, DVs in experiments or variables in Pearson correlations

• Examples: Fahrenheit temperature, IQ scores, Likert-type scales

Ratio data

• An interval scale with an absolute meaningful zero

• Direct comparisons (ratios) possible:

• 160 lb person weighs twice as much as an 80 lb person

• Often, DVs in experiments or variables in Pearson correlations

• Examples: height, income, cholesterol level, weight, reaction time, age

Importance of Scales - nominal

• In t-tests and ANOVAs, you must have a categorical/discrete IV (nominal).

• If you have a continuous IV, you can use regression techniques instead!

Importance of Scales - Nominal/Ordinal Variables as DVs:

• non-parametric tests like

  • Chi-Square

  • Wilcoxian test

  • Mann-Whitney test

  • Spearman correlation
    

Importance of Scales - Interval/Ratio Variables as DVs

• allow for use of parametric statistical tests like

  • t-test

  • ANOVA

  • multiple regression

  • etc.

Importance of Scales - 2 continuous (interval/ratio) variables

• can be correlated with Pearson correlation test.

• If one of two variables are categorical, you use other types of correlations (e.g., Spearman Correlation)

Discrete/Categorical variables

• Separate, indivisible categories; items classified into non-overlapping categories; typically nominal/ordinal variables

  • e.g., pregnancy status; drink size, number of students, gender, conditions

Continuous variables:

• Infinite number of divisions between categories; gradual continuum; typically interval/ratio variables

  • e.g., height in inches, voice pitch, income, blood pressure

Discrete vs. Continuous variables - Not always crystal clear!

• Many can be transformed into either type

• Can depend on whether variable is an IV, quasi-IV or DV

  • Self-esteem or Blood pressure

Operational Definitions

• Set of procedures used to measure or manipulate the variable

• Some operational definitions are better than others; relates to construct validity

  • Examples: Generosity, Hunger, Pain, Typing Skill, Self-
    Esteem

Operational Definitions Benefits

• Discuss abstract concepts in more concrete terms that are easier to analyze

• Can indicate that variable is too vague to study

• Helps others replicate your study

Between-Subjects Design - Advantages

• Minimal time required per participant

• No practice or fatigue effects

• Useful when made impossible to participate in all experimental conditions

Between-Subjects Design - Disadvantages

• Harder to find statistical differences between conditions, given the large degree of between-group variance

• More participants are needed per condition; increases experimenter’s total time/effort spent running studies

Within-Subjects Design - Advantages

• Fewer participants needed

• Extremely sensitive to statistical differences

• Person is their own control group; error from individual differences minimized

Within-Subjects Design - Disadvantages

• More time consuming per participant

• Participants might figure out what you are studying

• Order/carryover effects (from IV)

• Practice or fatigue effects

Randomized Counterbalancing

Order of presentation of conditions experienced by participant is systematically varied, allowing us to test for “order effects”

• SUPER thorough, but only works if small # conditions (<4)

Randomized Counterbalancing - 2 conditions

2 conditions = 2 orders (AB or BA)

Randomized Counterbalancing - 3 conditions

3 conditions = 3 * 2 * 1 = 6 “orders”
(ABC, CBA, ACB, BCA, CAB, BAC)

Randomized Counterbalancing - 4 conditions

4 conditions = 4 * 3 * 2 * 1 = 24
“orders” (yikes!)

Latin Square Designs

• A solution to randomized counterbalancing

• Fewer orders to test, but each order still occurs at least one time in the start, middle, and end of the experiment.

Latin Square Designs - Downside

order effects not completely eliminated

Factorial ANOVAs

The most basic of these between or within-subject designs apply for studies in which there is only one IV you are considering at a time, but sometimes you have multiple IVs, not just one.

Within-Subjects Factorial ANOVAs

2+ IVs are BOTH within-subjects variables

Between-Subjects Factorial ANOVAs

2+ IVs are BOTH between-subjects variables

Mixed-Subjects Factorial ANOVAs

1 IV is between-subjects and at least 1 IV in the same study is within-
subjects

Mixed-Subject Designs

• You have 2 IVs (can be mixture of quasi and manipulated): At least one of your IVs is “within-subjects” and at least another one of your IVs is “between-subjects”

Mixed-Subject Designs - Example

A pretest is given before the experimental manipulation is introduced to make sure groups are equivalent at the beginning of the experiment