1.
Internal Validity
Definition: The degree to which changes in the dependent variable (effect) can be
attributed to the independent variable (cause), and not to other factors.
Threats to Internal Validity
1. Selection Bias
○ Occurs when differences exist between groups (experimental vs. control)
even before treatment.
○ Example: A smoking cessation intervention (SCI) study—20 volunteers chose
to stop smoking (experimental), while 20 refused (control). The experimental
group may have been more motivated from the start, biasing results.
2. History
○ Refers to outside events occurring during the study that influence the
dependent variable.
○ Example: Pregnant women attend more prenatal training after a teaching
program, but a published article about maternal complications during the
program may have influenced them (not just the teaching).
3. Maturation
○ Changes naturally occurring in participants during the study may affect
results.
○ Example: Malnourished children may naturally gain height/weight during
the study, regardless of intervention.
4. Testing
○ When pre-test exposure influences post-test results.
○ Example: Subjects score higher on post-tests simply because they became
familiar with test questions.
5. Instrumentation Change
○ Differences in results are due to changes in the measurement tool, not the
treatment.
○ Example: A sphygmomanometer becomes faulty, altering blood pressure
readings.
6. Mortality (Attrition)
○ When dropout rates differ between groups, biasing results.
○ Example: Participants with low pre-test scores drop out, artificially raising
group post-test averages.
2. External Validity
Definition: The degree to which study results can be generalized to other people, settings,
or situations.
Threats to External Validity
1. Hawthorne Effect
○ Participants change behavior because they know they are being observed.
○ Solution: Use control groups, placebo, and blind experiments.
○ Example: A student nurse performs procedures correctly only because the
researcher is watching.
2. Experimenter Effect
○ Researcher’s behavior (facial expressions, gender, clothing) influences
participants.
3. Reactive Effect of Pre-Test
○ Taking a pre-test sensitizes subjects, influencing post-test results.
4. Halo Effect
○ Researcher’s impression influences rating of participants.
○ Example: An “intelligent” student nurse is automatically given a high score
without fair assessment.
Control Methods
● Double-Blind Method: Neither subjects nor observers know group assignments or
objectives.
● Double Observer Method: Two observers independently record and compare
findings to reduce bias.
3. Construct Validity
Definition: The extent to which the measurement tool (e.g., questionnaire) truly measures
the intended concept or theory.
Threats to Construct Validity
1. Reactivity to Study Situation
○ Participants alter responses due to perceptions of the study.
○ Control: Blinding, using objective measures (e.g., hospital records), or
pre-intervention strategies.
2. Researcher Expectancies
○ Subtle cues from the researcher (verbal or nonverbal) influence
participants.
○ Control: Blinding, monitoring researcher behavior.
3. Novelty Effects
○ Behavior changes occur because treatment is new, not because of its actual
effect.
4. Compensatory Effects
○ Equalization: Others give control group “compensations” for not receiving
treatment.
○ Rivalry: Control group tries harder to “compete” with treatment group.
5. Treatment Diffusion / Contamination
○ Control and experimental conditions blur.
○ Example: Control participants gain access to intervention; smokers
misclassify themselves as non-smokers.
4. Statistical Conclusion Validity
Definition: The validity of inferences about relationships between variables, based on
proper statistical methods.
Threats
1. Low Statistical Power
○ Small sample sizes make it hard to detect true relationships.
○ Solution: Larger samples, accurate tools, control confounding variables,
powerful statistical tests.
2. Restriction of Range
○ Limiting variability in data reduces the ability to detect relationships.
○ Example: Studying depression in elders <80, when most under 80 are not
depressed → no variability, so no relationships are detected.
3. Unreliable Implementation of Treatment (Intervention Fidelity)
○ Treatment not delivered consistently or participants not fully adhering.
○ Example: Music therapy study—participants must perceive “soothing” vs.
“jarring” music as intended; otherwise, treatment effect is unclear.
5. Trustworthiness in Qualitative Data
Lincoln & Guba (1985) identified 5 criteria:
1. Credibility – Confidence in truth of data.
○ Prolonged engagement
○ Persistent observation
○ Triangulation (multiple sources/theories)
○ Peer debriefing, member checks
○ Search for disconfirming evidence
○ Comprehensive and vivid field notes
2. Dependability (Reliability) – Stability of data over time.
○ Stepwise replication (split-half groups of researchers)
○ Inquiry audit (external review of data and findings)
3. Confirmability – Neutrality of data (free from researcher bias).
4. Transferability (Generalizability) – Findings can apply to other settings.
○ Requires thick description of setting, participants, and processes.
5. Authenticity – Fair and faithful representation of participants’ voices.
6. Statistics
Definition
● Branch of mathematics used to summarize, organize, present, analyze, and
interpret numerical data.
● Helps in research design, prediction, forecasting, and decision-making.
Types of Statistics
1. Descriptive Statistics
○ Summarize and describe data.
○ Tools:
■ Frequency Distribution (scores arranged highest → lowest)
■ Graphic Presentation (charts, graphs)
■ Measures of Central Tendency (mean, median, mode)
2. Inferential Statistics
○ Make predictions or generalizations from sample to population.
○ Purposes:
■ Estimate population parameters
■ Sampling error: Difference between sample data and
population data.
■ Sampling distribution: Theoretical distribution of infinite
samples.
■ Sampling bias: Sample not representative (e.g.,
non-probability sampling).
■ Test Null Hypothesis
■ Steps:
1. State research hypothesis.
2. State null hypothesis.
3. Choose appropriate statistical test.
4. Determine significance level.
5. Accept/reject null hypothesis.
Normal Curve Distribution
● Theoretical distribution of all possible values in a population.
● Symmetrical, bell-shaped.
Hypothesis Testing
● Non-directional Hypothesis: Extreme values may occur in either tail.
● Directional Hypothesis: Predicts specific direction of difference.
Tests of Significance
● One-tailed Test: Tests directional hypothesis (extreme values in one tail).
● Two-tailed Test: Tests non-directional hypothesis (extremes in both tails).
