Research Design and Statistics in Speech-Language Pathology

All professional practices are based on basic and applied research.
Current emphasis in speech-language pathology (SLP) is on evidence-based practices (EBP).
- Understanding the philosophy of science, research methods, data analysis, and conclusions is essential.
SLP and audiology are clinical sciences that rely on research evidence across disciplines.
Clinicians are expected to critically evaluate research to choose effective assessment and intervention methods.
Coursework in research design, statistics, and EBP is assumed to have been completed by readers, and this chapter aims to refresh this knowledge.

Science originated from curiosity and the quest to comprehend the world.
It is rooted in philosophical inquiry about truth, with empirical philosophy underpinning modern science.
Definition of Science: A philosophy of events and nature valuing evidence over opinion.
Research: The action phase of science, where scientists systematically ask and answer questions.
Science is conceptual and philosophical, while research follows a methodological framework.
Two key philosophical tenets of science are:
- Empiricism:
- States observations must be verifiable through sensory experience (sight, hearing, taste, touch, smell).
- Valid scientific knowledge relies on this type of evidence.
- Determinism:
- Asserts that events are caused by preceding events; they do not occur randomly.
- Science aims to discover event causes.
Goals of science include:
- Describing natural phenomena.
- Understanding and explaining phenomena, particularly cause-effect relationships.
- Predicting future events.
- Controlling phenomena through understanding their causes.

Scientists may explain events using:
- Inductive Method:
- An observation-driven approach where experiments lead to theories.
- Deductive Method:
- An explanation-driven approach where theories are proposed and then tested.
Hypothesis Formation:
- A hypothesis is a specific, testable statement derived from a broader theory.
- Types of Hypotheses:
- Null Hypothesis (H0): Assumes no relationship between variables.
- Alternative Hypothesis (H1): Assumes a relationship exists.
- Example scenario: Hypothesizing that stress increases stuttering.

Data collection involves systematic observations and measurements related to phenomena.
Scientific data is classified into:
- Qualitative Data: Non-numerical descriptions (e.g., “client has a severe articulation disorder”).
- Quantitative Data: Numerical descriptions (e.g., “client omitted 75% of word-final phonemes”).

Validity: The degree to which an instrument measures what it claims to.
- Types include:
- Predictive Validity: Accuracy of predictions on related tasks.
- Concurrent Validity: Correlation with established measures of validity.
- Construct Validity: Consistency with theoretical constructs.
- Content Validity: Assessment of the full range of the skill measured.

Experimental Research: Investigates cause-effect relationships through manipulation of variables.
Types of Designs:
- Group Designs: Compares group averages (experimental vs control).
- Single-Subject Designs (SSDs): Focuses on individual performance with few participants (common in treatment evaluations).
Definitions:
- Independent Variable: Manipulated by the researcher.
- Dependent Variable: Measured in response to changes in the independent variable.
- Extraneous Variables: Unwanted variables should be controlled in experiments.

Group designs analyze performance averages across groups.
Randomization: Essential for participant selection to minimize bias.
Control Group: Does not receive treatment, critical for confirming treatment efficacy.

Evaluates the relative effectiveness of multiple treatments across different groups.

Allow detailed analysis of individual performances without group averages.
Types of SSDs:
- ABA Design: Measures skills pre- and post-treatment, plus withdrawal phase.
- ABAB Design: Adds a final reinstatement of treatment phase to end on treatment.
- Multiple-Baseline Design: Staggered treatment application across subjects or settings to demonstrate treatment effects.

Internal Validity: Ensures data reflect true cause-effect relationships without contamination from extraneous variables. Potential threats include instrumentation errors, participant maturation, testing effects, and selection biases.
External Validity: The extent to which findings can be generalized to wider populations. Influenced by the Hawthorne effect and pretest sensitization effects.

EBP necessitates assessment of treatment research evidence, integrating clinician expertise and client preferences.
Classifications:
- Class I: Randomized clinical trials.
- Class II: Well-designed but non-randomized studies.
- Class III: Based on expert opinions and case studies.
Alternative classifications emphasize all valid research regardless of design.

Statistics involve organizing, summarizing, and analyzing data to draw conclusions.
Measures of central tendency and variability assist in data interpretation.
Different measurement scales (nominal, ordinal, interval, ratio) apply to research data and dictate the type of analyses possible.

Key Definitions

Statistic: A measure derived from a sample.
Parameter: A measure derived from a population.
Experimental Design: Involves manipulation of the independent variable to confirm its effects on the dependent variable, ruling out confounding factors.