Research Exam 1

Research

Systematic process of collecting and analysing data to increase our knowledge of a phenomenon with which we are concerned or interested.

Scientific Method

The principles & processes used in science for data collection & analysis

Steps of the Scientific Method

1.Pose a question
2.Seek answers
3.Develop a hypothesis
4.Conduct experiments
5.Analyze data
6.Draw conclusions
7.Disseminate findings

Experimental Research

-Explains phenomena through control of variables
-Aims to determine cause and effect

Descriptive Research

- Characterizes phenomena as they exist
-No real concern for cause and effect

Basic Research

- Driven by curiosity to expand knowledge
-Utilizes scientific theory
-Provides scientific validity
-Goal is to make something new

Applied Research

-Driven by desire to create solutions & solve problems
-Utilizes practice and experience
-Provides ecological validity
-Goal is to make something that already exists better

Laboratory

-Occurs in an artificial environment
-Utilizes experimental design
-Allows for significant experimental control and manipulation of variables

Field

-Occurs in a natural setting
-Utilizes primarily observational designs
-Normally includes few if any experimental controls or manipulations

Quantitative

•Numbers, numbers!
•Single reality/truth
•Seeks cause & effect
•Goal is to reduce error and bias
•Objectivity is primary
•Uses statistics
•Uses scientific methods

Qualitative

•Words, words!
•Multiple realities/truths
•Seeks understanding
•Goal is to understand multiple perspectives
•Subjectivity is primary
•Uses descriptions
•Uses constant comparison & revision

Influences Confounding the Research Process

Tradition, superstition, authority, bias

Reporting Scientific Findings

Introduction, Methods, Results, Discussion, References

Peer Review Process

Process by which a manuscript is evaluated to determine quality and appropriateness for publication

Unblinded Review

Authors and reviewers both know each other's identities

Single-Blinded Review

Reviewers know authors' identities but authors do not know the reviewer's identities

Double-Blinded Review

Neither authors nor reviewers know each other's identities

Evidence-Based Practice

The fusion of the latest research, clinical expertise, and patient values to guide care decisions.

Variables of EBP

1. Scientific Evidence: current, best, applicable.
2. Clinician Experience: expertise over time
3. Client Goals: values, perspectives

Steps of EBP

1. Formulate a specific question
2. Search and review relevant literature
3. Evaluate the research and consider all options
4. Make a decision and incorporate it in practice
5. Evaluate the process, decision, and outcomes

Evaluation System Considerations

Criteria should be pre-established, using a grading system such as one developed by professional organizations

Systematic Reviews

Thoroughly synthesizes multiple relevant studies on a single question, providing a high-level overview with strict inclusion criteria, surpassing traditional expert reviews in rigor.

Meta-analyses

A systematic review method that combines data from various studies into a common factor, allowing for statistical analysis to determine effect size. It adheres to strict guidelines and provides pooled data for interpretation.

Primary Scientific Literature

Original research, generated by the authors. Ex. Original research, conference papers, technical reports, theses, dissertations

Secondary Scientific Literature

Research from others gathered and summarized by the authors. Ex. Review articles, meta-analyses, textbooks

Impact Factor

Average number of times a journal's articles are cited over a period of time. Formula: # of Citations / # of Publications

Dependent Variable

Variable measured to assess impact or response to a manipulation. Also known as a response or output variable.

Independent Variable

Variable manipulated by the researcher or nature. Also known as an explanatory or predictor variable.

Quantitative Variable

Describe variables that are measured numerically

Categorical Variable

Describe variables that cannot be measured easily with numerical values

Levels of Measurement: Nominal Variable

Without rank or order

Levels of Measurement: Ordinal Variable

With rank/order not necessarily evenly spaced

Levels of Measurement: Interval Variable

With rank/order and evenly spaced but without a true zero point

Levels of Measurement: Ratio Variable

With rank/order, even spacing, and true zero point

Extraneous variables

Uncontrolled factors in research that can disrupt the link between the independent and dependent variables.

Descriptive Research

Provides initial information in emerging areas without establishing cause and effect.

Observational Research

Longitudinal studies tracking natural groups over time to establish cause-and-effect relationships.

Experimental Research

Manipulates exposure to measure outcomes, establishing cause and effect.

Case Study

Descriptive Research- evaluates phenomenon for an individual case at a point in time

Correlational Study

Descriptive Research- evaluates a sample to assess association between variables at a point in time

Case Series

Extends the case study over time

Epidemiological Cross-Sectional Study

Variation aimed at identifying risk factors and related outcomes

Cohort Study

Observational Research- Evaluates risk prospectively by establishing present exposure and noting future outcomes. Yields the Relative Risk statistic.

Case-Control Study

Observational Research: Evaluates risk retrospectively by establishing current outcomes noting prior exposure. Yields the Odds Ratio statistic.

Randomized Controlled Trials

Represents a true experiment by way of exerting controls to limit alternative explanations

Non-Randomized Intervention Studies

Any design that is experimental in nature but does not meet the standards of an RCT

Double Blind (In Terms of Treatment)

Neither the participant or investigator are aware of which group is the treatment group

Placebo

Prevents participants from figuring out which group is receiving the intervention vs. the 'sugar pill'

Crossover

Allows each participant to serve as their own control

Random Error

Often from many directions & yielding net zero effect. Not typically reproducible & so 'unreliable.'

Systematic Error

Mostly in one direction & yielding a nonzero effect. Typically reproducible & so 'reliable'

Hierarchy of Evidence Pyramid

1) Ideas: anecdote, expert opinion
2) Bench Science: animal models, mechanistic work
3) Epidemiology: descriptive, observational
4) Controlled Research: clinical trial, RCT
5) Synthesis: meta-analysis, systematic review

Design Approach

Factors impacting study design:
Existing data, Risk and ethics, Resources, Timeframe, Sample, and measures
Balance between preference and necessitity

Frequentist Approach

- Focuses on probability that an observation is 'real'
- Forms the basis of hypothesis testing and p-values

Magnitude Approach

- Focuses on the size of observed differences/similarities
- Forms the basis for effect sizes and confidence intervals

Individual Approach

- Focuses on the individuals not group averages
- Forms the basis of individual odds for specific outcomes

Null Hypothesis

- Assumes alternate finding is due to chance/sampling error
- States there is 'nothing here'

Research Hypothesis

- Assumes alternate finding is due to experimentation
- States that 'we found something'

Type I Error

Null is true and you think the null is false

Type II Error

Null is false and you think the null is true

Number of Dependent Variables

1 = Univariate
2+ = Multivariate

Number of Independent Variables

1 = Single factor design
2+ = Factorial design

Number of Independent Variable Categories

2 = Typically allows for more basic analysis
3+ = Requires sophisticated analyses and follow-up

Between-Subjects

Different participants in each level

Within-Subjects

Same participants in each level

Ways to Manage Order Effects

- Crossover: participants serve as own control
- Randomization: uses 'draw out of a hat' approach; limits likelihood of order effects
- Counterbalancing: order is rotated among possible scenarios
- Balanced Latin Square: extension of counterbalancing with 3+ conditions

True Experimental

Aims for causality & internal validity; lots of design control; randomization; always prospective

Control Group

Design group receiving standard or placebo treatment

Single Blind (In terms of treatment)

Participants don't know if they receive experimental or placebo

Double Blind (In terms of treatment)

Participants and some/all study personnel don't know who receives what

Placebo Effect

Benefits to participants without receiving experimental intervention

Hawthorne Effect

Subjects performing better due to being observed

Inclusion Criteria

Guidelines that must be met to be included in a research study. Homogeneity = facilitates design control. Heterogeneity = facilitates generalizability.

Internal Validity

Improved with greater control, worsen real life applicability

External Validity

Worsen with greater control, improved real life applicability

Validity

Degree to which a tool measures what it is intended to measure Represents 'truth' or 'accuracy' in measurement

Reliability

Degree to which a tool measures in a stable manner Represents 'consistency' or 'stability' in measurement

Regression

- Provides a measure of variable association
- Produces a measure of the strength & direction of the relationship
- Based on interclass (b/w variables) correlation

Bland-Altman

- Provides a measure of variable agreement for highly correlated variables
- Produces a measure of the 'differences' between variables

Intrarater Reliability

One scorer; multiple time points

Interrater Reliability

Multiple scorers at one time point

Intraclass Correlation

Based on the relation with a single class/variable. Provides a measure of systematic variance.

Standard Error of Measurement

Describes how often to expect errors of a given size. Provides the 'standard deviation' of error.

Systematic Review vs. Traditional Review

Systematic Review:
- Objective, comprehensive assessment of a research question
- Examines all relevant research papers with strict inclusion criteria
- Highly structured and minimizes bias
Traditional Review:
- Subjective, selective evaluation of a research question
- Considers papers chosen by the author as most relevant
- Less structured and prone to bias

What is Meta-Analysis?

Quantitative review after a systematic review.
Mathematically analyzes data from multiple studies.

What is the Primary Outcome of Meta-Analysis?

Determines treatment effect.Results in an Effect Size (ES). ES can be expressed in various ways based on data type.

What is an Effect Size (ES)?

A statistical measure that quantifies the strength of a relationship or treatment effect in research.

Fixed Effects

Identifies the 'best' ES estimate

Random Effects

Identifies the 'average' ES estimate

Homogeneity

Similar study results; easy to combine; generally use random effects model

Heterogeneity

Different study results; harder to combine; generally use fixed effects model

SMD Effect Size

ES = (mean 1 - mean 2) / (average SD of means 1 & 2)

Effect Size interpretation

Differences (SMD): Small = 0.2; Medium = 0.5; Large = 0.8; Very Large = 1.3

Confidence Interval

It acknowledges the uncertainty around the Effect Size (ES) estimate by providing a 95% confidence interval.

Forest Plot

It visually represents individual study Effect Size (ES) values with 95% confidence intervals and calculates a model value from all studies.

Systematic Review & Meta-Analysis Steps

1. Develop a focused question
2. Develop a protocol
3. Review the literature
4. Identify studies
5. Abstract data
6. Assess risk of bias
7. Perform meta-analysis

Ethos

Habitual way of behaving

Ethics

Guidelines for morally acceptable conduct

Autonomy

The right to govern oneself

Beneficence

One should act to do good