Research Methods and Validity Flashcards

Comprehensive vocabulary flashcards covering research methods, validities, variables, sampling strategies, threats to validity, study designs, and epidemiological statistical formulas.

Quantitative research

Research that collects and reports data primarily in numerical form; characterized by larger samples for more generalizable results, specific hypotheses & variables.

Qualitative research

Focuses on the "why and how" by seeking in-depth, open-ended responses; often uses smaller samples & explores personal experiences.

Quantitative validities

The four types of validity evaluated in quantitative research: external validity, internal validity, construct validity, and statistical conclusion validity.

Population of interest

The specific group relevant to the research.

Internal validity

The extent to which we can draw cause-and-effect inferences between the IV and DV in a study.

Construct validity

The extent to which variables measure what they are supposed to measure.

Statistical conclusion validity

Concerns the best statistical treatment of data and the proper interpretation of the researchers' statistical conclusions.

Independent variable

The variable that is being studied.

Active independent variable

An independent variable that the researcher assigns to a subject in an experimental study.

Attribute independent variable

An existing characteristic of a subject used in an observational study.

Dependent variable

The outcome that measures the effect of the independent variable.

Extraneous variable

Any variable other than the IV and DV that is not being studied, but could affect the DV.

Confounding variable

An extraneous variable that relates to both the IV and the DV and makes it hard to determine the relationship; researchers often try to control for it.

Covariate variable

A variable that relates only to the DV; researchers want to account for it but cannot change or influence it.

Mediating variable

A variable caused by the IV that influences the DV; explains the relationship between the IV and DV.

Moderating variable

A variable that affects the strength or direction of the relationship between the IV and DV.

Hypotheses

Declarative predictions that provide explanation for an observed event.

Correlational language

Terms like "relates", "associates", and "predicts" that describe one variable changing alongside another.

Causational language

Terms like "difference" and "improves/reduces" that describe one variable directly causing an effect in another.

Null hypothesis (H0)

The baseline assumption in statistical testing; usually posits "no effect".

Alternative hypothesis (H1)

What the researcher aims to support through the study; can be directional or non-directional.

Directional alternative hypothesis

An alternative hypothesis stating that the intervention will have a specific effect.

Non-directional alternative hypothesis

An alternative hypothesis stating that the intervention will have an unknown effect.

Reject the null hypothesis

Conclusion indicating there is a possibility that the treatment has an effect.

Fail to reject the null hypothesis

Conclusion indicating there is insufficient evidence that the treatment worked.

Boolean operators

Search terms AND (narrows), OR (expands), and NOT (excludes) used in search strings to refine the scope of a search.

Nesting

Using parentheses to group words and boolean operators in search strings (e.g., anxiety AND (reasons OR causes OR factors)).

Supernest

A search expression containing at least one nest and other additional search information.

Truncation

Using asterisks in search strings to retrieve variations of a word root (e.g., *child = childhood, childbirth, childproof, etc.).

Double quotes (search technique)

Grouping a phrase between double quotes to search for an exact whole phrase (e.g., "Public Health Policy").

Search tags

Indicators that tell a search engine where to focus its attention when looking for information.

Tag [Ti]

Search tag specifying the title field.

Tag [Tiab]

Search tag specifying the title and abstract fields.

Tag [MeSH]

Search tag specifying medical subjects/headings.

External validity

The extent to which we can generalize findings to real-world settings/general population.

Target/theoretical population

The segment of people that researchers are interested in studying.

Accessible sample (sampling frame)

Who could potentially be sampled; researchers rarely have good control over this.

Selected sample

Who researchers sample from the accessible population; researchers have much more control over this.

Actual sample

Those who actually participate in the study; researchers have little control over this.

Population validity

How reasonably we can generalize the findings from the sample in our study to the general population.

Ecological validity

How reasonably we can generalize the findings from our study to real-world situations.

Case series / case reports

Observational designs used for observing one or many patients with a specific condition.

True experimental design (RCT)

A study design where the IV is actively assigned by the researcher and subjects are randomly assigned.

Quasi-experimental design

A study design where the IV is actively assigned by the researcher, but subjects are not randomly assigned.

Observational study design

A study design involving an attribute IV and no random assignment of subjects.

Cohort study

An observational study design that starts with an exposure variable to determine an outcome.

Case-control study

An observational study design that starts with the OUTCOME in the PRESENT to determine the exposure in the past.

Cross-sectional study

An observational study design where exposure and outcome are collected at the same time.

Causal research

Research that aims to define the cause/effect relationship.

Causal research pyramid

Levels of evidence ranking study design from strongest to weakest; higher levels control extraneous variables better, while lower levels are less effective in determining causality.

Animal studies (pyramid rank)

Non-human research ranked 9th on the causal pyramid; often low-cost and fast, but with limited causal applicability to humans.

Case report / case series (pyramid rank)

Observational research ranked 8th on the causal pyramid; detailed account of diagnosis/treatment/outcome of patient(s), useful for novel phenomena but not causality due to small random samples.

Cross-sectional study (pyramid rank)

Observational research ranked 7th on the causal pyramid; measures DV and IV at the same point in time, hinting at causality without establishing it.

Case-control study (pyramid rank)

Observational research ranked 6th on the causal pyramid; compares those with a condition to a control group without it, looking back at contributing factors.

Cohort study (pyramid rank)

Observational research ranked 5th on the causal pyramid; follows groups exposed to a risk factor over a long period to determine likelihood of developing a condition.

Quasi-experimental study (pyramid rank)

Experimental research ranked 4th on the causal pyramid; active IV but no random assignment, limiting causality establishment due to uncollected confounding variables.

True experimental (RCT) (pyramid rank)

Experimental research ranked 3rd on the causal pyramid; active IV and randomly assigned groups to minimize confounding variables, serving as the gold standard for causality.

Systematic reviews (pyramid rank)

Synthesized research ranked 2nd on the causal pyramid; examines existing studies on topics to draw evidence-based conclusions, dependent on quality of reviewed studies.

Meta-analysis (pyramid rank)

Synthesized research ranked 1st (top) on the causal pyramid; type of systematic review combining data through statistical techniques, providing the strongest evidence.

Probabilistic sampling

Non-biased sampling where everyone in the population has a chance of being selected.

Non-probabilistic sampling

Methods for choosing subjects that do not use a random selection process; used when probabilistic is not feasible or for qualitative studies.

Simple random sampling

A sampling technique where everyone in the population has an equal chance of being selected.

Systematic random sampling

Selects a sample from an ordered arrangement of the population by randomly selecting one of the first n individuals and choosing every nth individual thereafter.

Stratified random sampling

Separation of the target population into different groups/strata based on a common characteristic, and selecting samples from each stratum.

Equal proportions stratified sampling

Uses different percentages across strata, resulting in an unequal number of people in groups; often good for surveys.

Different proportions stratified sampling

Uses the same percentage across strata, resulting in an equal number of people in groups; good for experiments.

Oversampling

When a researcher intentionally over-represents one or more groups to ensure comparison of equal numbers of subjects.

Cluster sampling

A sampling technique in which clusters of participants that represent the population are used; often used for efficiency with limited resources.

Primary validity goal of experimental designs

Experimental designs specifically attempt to maximize internal validity.

Key differentiator of experimental designs

What separates experimental designs from observational designs is the presence of an active IV.

Disadvantages of experimental designs

Cons include being costly and time consuming, high internal validity but lower external validity, and participant variability among groups.

Pre-test / baseline test

Measurement taken before starting an experiment to assess where things stand at the beginning.

Control group

The group not receiving the treatment/intervention (may receive placebo or standard care) to help account for external factors or placebo effect.

Random assignment of subjects

Applies only to true experimental studies; allows researchers to more confidently say an observed effect is due to the intervention.

Matching / matched pairs

Matching subjects on characteristics relevant to DV, then assigning them to IV levels; important for small sample sizes or confounding variables.

Blinding

Procedure where experimental and control groups do not know which group assignment they received.

Double-blinding

Procedure where subjects and researchers both do not know group assignments.

Within-subjects design

Exposing the same participant to multiple conditions over time; good for working with limited resources.

Longitudinal study

Collecting data on separate intervals of time across subjects; often costly and difficult to maintain participant engagement.

Crossover design

Design where after pre-test/intervention/post-test, the experimental group becomes the control group and vice versa, controlling for individual differences.

Regression towards the mean

The tendency for extreme or unusual scores or events to fall back (regress) toward the average; statistical threat to validity.

Attrition bias

Occurs when participants drop out of a long-term experiment or study; worse when one group has a much higher dropout rate than the other.

Allocation bias

Systematic difference between how subjects are selected or assigned for one group versus another group; also known as selection bias.

Testing bias

Possible fatigue or carryover effect from pre-test to post-test; more likely with short intervals between tests.

Maturation threat

When participants naturally change as a function of time rather than because of the IV; older and younger people are more prone.

Performance bias

When subjects modify behavior after finding out their group assignment, or when caregivers deliver different care due to expectations.

History bias

External event occurring before or between pre-test and post-test that alters response (e.g., 9/11, COVID pandemic).

Detection bias

Occurs when members of the research team have a preconceived notion of treatment efficacy, or fail to follow protocol.

Biases counteracted by random assignment

Regression to the mean, attrition bias, and allocation bias.

Biases counteracted by blinding

Attrition bias, performance bias, and detection bias.

Biases counteracted by control groups

Regression to the mean, testing bias, maturation threat, and history bias.

Biases counteracted by quality assurance

Attrition bias, performance bias, and detection bias.

Observational study design strength order

Types of observational studies ranked from weakest to strongest: 1. Case report/case series, 2. Cross-sectional, 3. Case-control, 4. Cohort.

Prospective cohort study

Starts by separating people by EXPOSURE in the PRESENT, and determines differences in outcome in the future.

Retrospective cohort study

Starts by separating people by EXPOSURE in the PAST, then determines differences in outcome in the present.

Analytical data

Data used to compare groups (e.g., assessing if physically inactive adults have higher obesity prevalence than active adults).

Descriptive data

Data providing prevalence estimates (e.g., how many adults have diabetes right now).

Relative risk (RR)

Incidence of exposed individuals divided by incidence in non-exposed individuals; calculated only for COHORT studies.

Odds ratio (OR)

Odds that an exposed person develops disease divided by odds that a non-exposed person develops disease; used for cohort or case-control studies.

2x2 table flow for cohort studies

Start on the left with exposure (IV) and move right toward the outcome.