Variable - Elements of a Claim

  • variable: something that varies/can take on different values(quantity = numbers /quality = colors/labels/words)/changes

    • NOT CONSTANT

    • examples:

      • ex. you are an extroverted person - not a variable

      • ex. a student used active recall for studying for studying and received an A for a class - NOT VARIABLE (CONSTANT)

        • scientific if: multiple study strategies related to grades

      • ex. the customers who had sweet tea reported a high level of satisfaction - not variable

        • scientific if: compared to different drinks

      • ex. after doing regular exercise, people’s stress decreased - ish?

        • compares stress before and after exercise, high to low, could be clearer

  • conceptual vs operational variables:

    • conceptual: an abstract label for a concept being studies

      • also called constructs in psychology

      • ex. extroversion, stress, study strategy

    • operational: how do i observe it?

      • how a variable is actually represented in a study through manipulation, observation, measurement, etc.

      • more tangible, and often quantifiable

        • examples:

          • ex. self report scores on a questionnaire

          • ex. heart rate variability

          • ex. being asked to create a story about the concepts learned from class (or not)

      • manipulation: researchers actively create or control what the participants do or experience

        • manufactured/induced/created by researcher

        • variable being manipulated is called the independent variable

        • a study that involves at least one manipulated variable is called an experimental or quasi-experimental study (allows for causation statements)

          • quasi = not completely random (no random assignment)

            • may not be feasible for total random assignment, but still involves variable manipulation

          • examples:

            • ex. intake of drug or placebo controlled by researcher

            • ex. doing a task with or without a specific time limit instructed by researcher

            • ex. reviewing profiles with different photos, phots selected and presented by researcher

          • limitation: ethics, bias, are the variables/results “natural” and represent every day life vs are they artificial and yielding artificial results

          • strengths: can establish causation

      • observation: researchers observe naturally occurring behavior or other types of evidence

        • an observed variable can be used in any type of study

          • ex. dependent variable in an experimental study

          • ex. a predictor or outcome variable in a non-experimental study (completely naturally occurring variables, no manipulation, no independent/dependent labels)

        • examples of predictor/outcome:

          • ex. time spent of physical activity every day - naturally occurring (predictor)

          • ex. level of activation of a specific brain area (outcome)

          • ex. messiness of space after a visit (indicator)

        • is evidence collected from observation objective?

          • subjectivity in relating evidence to conceptual variables

      • subjective measures: participants use opinion to report on variables related to themselves or others

        • a subject measurement can be used in any type of study, similar to observed variable

        • may be the only option when a variable is related to someone’s internal mental process

        • exmples:

          • ex. self report on satisfaction level on a scale

          • ex. others’ evaluation of a worker’s performance on a scale

          • ex. clinician assessment of someone’s mental health on a scale

  • causation vs correlation - types of claims

    • frequency claims: focused on 1 variable at a time

      • examples:

        • ex. frequency or percentage of people doing one thing

        • ex. mean and standard deviation of the data on a variable

    • correlation: the 2 or more variables are related

    • causation : one (or more) variable causes another variable (or multiple variables)

    • these 2 claims can be used to described hypotheses (specific predictions) of findings

      • examples:

        • ex. warm (vs cool) room temperature will harm cognitive performance

          • variables: room temp, cognitive performance - causation statement

        • ex. individualistic cultural value is associated with pro social behavior

          • variables: cultural value, pro social behavior - correlation statement ex

        • ex. a new medicine vs placebo will decrease the symptoms lf a neurological disorder

          • variables: meds, symptoms - causation statement

        • ex. level of patience is tied to number of traffic violations

          • variables: patience, traffic violations - correlation statement

  • correlation descriptors:

    • stating the claim: if the predictor variable is related/associated/correlated to the outcome variable

      • non directional: stating 2 variables are related

        • social media activity and stress stress are related: predictor = social media activity, outcome = stress

      • directional positive: as 1 variable increase another increase, vice versa

        • performance in college and income are positively related: predictor = performance in college, outcome = income

      • directional negative: as 1 variable increases another decreases

        • sleep quality and fatigue are negatively correlated: predictor = sleep quality, outcome = fatigue

      • evidence needed for correlation: scatter plots

        • requires observations/measurements of both variables on same group of individuals

        • statistical analysis

          • direction: positive/negative

          • magnitude/strength: between -1 and 1

            • (-)1 = strongest

            • 0 = no correlation

          • test of statistical significance

  • causation = evidence needed:

    • covariation:

      • need evidence on both IV and DV from same people

      • need to use statistical methods to examine the relationship

    • temporal precedence:

      • need evidence to support ordering of events: change in IV → change in DV

      • active control of IV by research (experimental design)

      • following the same people over time (longitudinal design)

    • no other explanations:

      • need to rule out confounds

      • active control of potential confounds and ensure the IV is the only variable being manipulated (experimental design)

      • observing/measuring potential confounds and control for them statistically (any design)

    • statement: the IV causes the change in DV

  • tips for writing clear claims:

    • both (or all) variables are clearly represented

      • clear names/labels for all variables

      • clear levels (values, groups, etc) of all variables

        • example:

          • warm room = not variable

          • room temp = variable

    • the relationship or difference is clearly stated

      • clear about correlation or causation

        • correlation: patience and number of traffic violations are related

        • causation: patience affects the number of traffic violations

  • validity: evidence → claims

    • correlation ≠ causation

    • inference: a conclusion made based on evidence but not directly stated in the evidence

    • validity: the quality of the inference made

      • internal validity: the quality of a cause and effect claim made by a study

        • stats, covariation, temporal precedence, random assignment, manipulation/control

      • external validity: the generalizability of a claim made by a study.

        • based on: sample type, size, representativeness, etc.

    • validity of an operational variable:

      • score/type on an operational variable (observation + manipulation)

      • variable measured/represented accurately

      • not the validity of the whole claim but is important for the validity of a claim

        • different types of evidence:

          • reliability: (necessary but insufficient for validity) based on data

          • validity: based on judgment and date

  • reliability of an operational variable:

    • reliability: consistency (not agreement)

      • accounts for free form random error: based on the assumption that errors are random (not correlated with any systemic factors)

      • assume we are measuring same group on one variable:

        • test retest: scores from different times should be consistent (not identical but within similar ranges)

        • inter-rater: scores from different raters should be consistent (various “trained” researchers/interviewers)

          • interviews should align

        • internal consistency: scores on different items should be consistent

          • commonly seen in peer reviewed research (method section)

            • asking the same factor differently should yield similar results

        • cronbach’s a (alpha coefficient) (need > .70)

  • evidence for validity summary:

    • face and content validity: can we infer the construct of interest based on the content of the measure

      • relies on judgement

    • construct and criterion related validity: can we infer people’s standing on the construct or an outcome based on the scores on the measure

      • relies on statistical analysis