Week 3: Experimental design

Variable

Anything that varies. Quantity that can change; usually used to refer to a measure of a phenomenon.

Independent variable (IV)

Variable that the experimenter manipulates as a basis for making predictions about DV.

Dependent variable (DV)

Variable that is measured or recorded in an experiment (outcome).

Experimenter or manipulated variables

Manipulated by the researcher - allocation to experimental group, such as timing/ dosage.

Participant/ subject variables

Not manipulated - gender, university course.

Confounding variables

Variables we know might affect our DV (or IV) that we are able to control/ correct for.

Extraneous variables

Variables that potentially influence results but are not of direct interest to our research.

• Can relate to the subject, experimenter, environment, etc.  

• E.g. Interaction with the researcher, medical conditions 

Continuous variable

Can take any value within a given range. Doesn't change in discrete jumps.

E.g., Temperature; levels of anxiety

Discrete variable

Can take on only certain discrete values within the range. Has to be a whole number.

E.g., number of cars owned; number of children in a family.  

Categorical variable

The value that the variable takes is a category. More limited.

E.g., gender; occupation; ethnicity

Dichotomising variables

Researchers sometimes convert continuous or discrete variables into 2 categorical variables.

E.g. splitting age into 2 categories, i.e. old vs young, so it is easier to compare 

Randomised controlled trials (RCTs)

Considered the gold standard of research. Most rigorous form of research.

• Used to measure the effect of an intervention by randomly assigning individuals to intervention or control group.  

• Participants and researcher are blinded to the condition. -double-blind: prevents investigator bias 

RCT example

The Women's Health Initiative Memory Study (Shumaker et al., 2003).

A randomised, double-blind, placebo-controlled clinical trial.  

• 4532 postmenopausal women free of dementia, aged 65+ years.  

• Randomly assigned to take either:  

• 1 daily tablet of estrogen plus progestin (n = 2229) 

• Matching placebo tablet (n = 2303). 

True experiments

Laboratory-based and fully controlled experiments that include experimental manipulation, standardised procedures, and random allocation of participants to conditions.

True experiment example

Inducing preschool children's emotional eating: relations with parental feeding practices (Blissett, Haycraft & Farrow, 2010).

Children randomly allocated to a control or negative mood condition – received a sticker after completing jigsaw vs jigsaw with missing piece (only receive sticker if completed)

Quasi experiments

• Random assignment into treatment and control groups: non-equivalent groups, random assignment unable to take place 

• Full control over the independent variable (the manipulated variable)  

• Sometimes it’s not feasible or ethical to do true experiments 

Quasi experiment example

A quasi-experimental study of maternal smoking during pregnancy and offspring academic achievement (D'Onofrio et al., 2010).

• Maternal smoking during pregnancy cannot be manipulated (ethically) - found no differences 

Correlational studies

Studies used to determine if one factor is related to another, involving non-manipulated variables.

Correlational study example

Szinay et al. (2019) explored associations between self-esteem, smoking, and excessive alcohol consumption with 187,398 participants.

Questionnaires

Commonly used to collect data in correlational studies and to objectively measure a particular concept.

Psychometrics

The area of study concerned with the theory and technique of psychological measurement.

• Diagnosis or screening for clinical purposes  

• E.g. Beck Depression Inventory, Eating Disorders Examination Questionnaire, Brief Psychiatric Rating Scale etc. 

Causal relationships

Relationships where one thing causes another, as opposed to merely measuring two variables.

RCT

Randomized Controlled Trial, a type of true experiment.

Within-Subjects Design/repeated measures

A research design where the same participants are used in every condition of the independent variable.

Evaluation of a within-subjects design/repeated measures

+Equivalent groups  

+ Need to recruit less participants  

-Order effects; practice effects; carryover effects  

-Attrition – drop-out rates 

-Equivalent stimuli 

-Demand characteristics 

Between-Subjects Design/independent groups

A research design where different participants are used in each condition of the independent variable.

Evaluation of a between-subjects design

+ No order effects  

+ Fewer demand effects/characteristics -can’t guess the aims of the study 

+Loss of participant from one condition only if there is drop out 

-Need to recruit more participants  

-More costly 

-Can’t control for individual factors as in W-S designs  

-If differences in variance of participants is too great, this limits statistical analyses 

Order Effects

Effects that occur when the order of conditions affects the outcome of the experiment.

Practice Effects

Improvements in performance due to repeated exposure to the same task.

Carryover Effects

Effects that occur when the effects of one condition carry over to another condition.

Attrition

The loss of participants from a study over time.

Counterbalancing

A method used in within-subject designs to eliminate order effects by varying the order of conditions.

Asymmetrical Order Effects

Order effects that have greater strength in one particular order of conditions.

→A possible solution is to complex counterbalance– ABBA

Random Allocation

The process of randomly assigning participants to different groups to ensure equal opportunity.

Pre-test

An assessment conducted before the main experiment to ensure groups are matched on relevant characteristics.

Mixed Design

A research design that includes a combination of between-subjects and within-subjects factors.

Briefly outline Stephens et al’s (2018) study

A study investigating whether swearing out loud increases physical power, using a within-subjects design.

 • Condition A: Participants suggests a swear word they might use in response to banging their head accidentally.  

• Condition B: A word used to describe a table.  

• Physical power on exercise bike measured. 

-> Found greater power for those who swore compared to those not swearing 

RCTs

Randomized Controlled Trials, a key feature of true experiments involving random assignment.

Equivalent Groups

Groups that are similar in all respects except for the treatment they receive.

Demand Characteristics

Cues that can inform participants about the purpose of the study, potentially influencing their behavior.

What is the difference between the sample and the population?

Population: a group that shares a common set of characteristics; the wider group you wish to learn about  

Sample: The group selected from the population to participate in your research. 

WEIRD samples

Samples that are Western, Educated, Industrialised, Rich, and Democratic.

Mixed design/mixed measure study

• Sometimes our research designs include a combination of between-subjects factors and within-subjects factors. 

– One or more IV uses the same participants (within) 

– One or IV uses different participants (between) 

Overcoming WEIRD samples

Proposed ideas for authors, journal editors, and reviewers to ensure psychological science is representative.

Probability-based sampling

Everyone in the target population has an equal probability of being selected.

Non-probability sampling

Sample is not structured to approximate the population.

Simple random sampling

Every member of the population of interest has an equal chance of being selected for participation.

Systematic random sampling

Select every nth case from the population, where n is any number.

Stratified random sampling

Take a random sample from various sub-sections of the population.

Opportunity/convenience sampling

The most common sampling strategy, but the lowest credibility; whoever is available to take part.

Self-selecting sampling

Volunteers for research, common in experiments.

Online samples

Very common for questionnaires/surveys, also experiments, large data set.

Quota sampling *

A non-probability sampling method where researchers select participants to ensure specific characteristics are represented.

Haphazard sample *

Sample selected from population with no conscious bias, but likely not to be truly random.