Research Methods: Design and Analysis

26
Research Methods: Design and Analysis by Dr. James Lani

description

Research Methods: Design and Analysis. Covering the research cycle, research questions, operationalization of variables, literature review, research designs, sampling method, instrumentation, data collection, validity, reliability, data analysis plan, and sample size

Transcript of Research Methods: Design and Analysis

Page 1: Research Methods: Design and Analysis

Research Methods: Design and Analysis

by Dr. James Lani

Page 2: Research Methods: Design and Analysis

Today’s Webinar

To schedule a consultation call 877-437-8622 or email [email protected]

• Why Do Research?• The Research Cycle• Research Questions and (Testable) Hypotheses

• Operationalization of Variables• Literature Review Resources• Research Designs• Sampling Method and Sample Assignment• Instrumentation• Method of Data Collection (Procedures)• Validity

• Internal Validity• External Validity

• Reliability• Data Analysis Plan• Sample Size• Q & A

Page 3: Research Methods: Design and Analysis

Why Do Research?

To schedule a consultation call 877-437-8622 or email [email protected]

OK, to get out of school….and also to examine questions we care about. We want to test a theory, evaluate a program or intervention, describe a phenomenon.

Does Violence on TV cause children to be more or less aggressive?Do opposites attract or birds of a feather flock together?Does treatment X predict academic achievement?Is there a difference on depression scores by group (exercise vs. not)?

What is True???Goal of research: describe, explain, and predict. To establish knowledge, you need methods and statistics.

Page 4: Research Methods: Design and Analysis

The Research Cycle

To schedule a consultation call 877-437-8622 or email [email protected]

LiteratureReview

MethodHypotheses and

Operational Variables

MethodDesign Study and

Collect Data

ResultsAnalyze data and

accept or reject Ho’s

Discussion

Page 5: Research Methods: Design and Analysis

Research Questions and (Testable) Hypotheses

To schedule a consultation call 877-437-8622 or email [email protected]

• Hypotheses: State relationships among variables (constructs). Typically in terms of Differences, Relationships, or Prediction.

• Keep research questions and hypotheses aligned (in number and language).

• Questions must be testable.• Keep it tentative: no proof—just support

or non-support

Page 6: Research Methods: Design and Analysis

Operationalization of Variables

(or how do you measure what you’re studying?)

To schedule a consultation call 877-437-8622 or email [email protected]

Constructs—objects that are not directly observable• Constructs are measured with variables• Should be quantifiable • (e.g., depression can be measured with a Beck

depression scale or by the observation of the number of times someone laughs)

Confounding Variables• Provide alternative explanation of response in the DV

Levels of Measurement: Nominal, Ordinal, Interval, and Ratio

Page 7: Research Methods: Design and Analysis

Literature ReviewResources

To schedule a consultation call 877-437-8622 or email [email protected]

See: UW—Madison—the writing centerUC Santa Cruz—UNW library—write a literature reviewWashington and Lee University—literature reviewStatistics Solutions’ Free membership—sample dissertation

Page 8: Research Methods: Design and Analysis

Research Designs

To schedule a consultation call 877-437-8622 or email [email protected]

EXPLORATORY

EXPERIMENTALDescriptive

Longitudinal

CAUSAL Cross-sectionalCorrelational Cohort

designs

Non-experimental

Sequential

Page 9: Research Methods: Design and Analysis

Research Designs

To schedule a consultation call 877-437-8622 or email [email protected]

Its about how you structure your inquiry

Page 10: Research Methods: Design and Analysis

Research Designs

To schedule a consultation call 877-437-8622 or email [email protected]

Experimental vs. Non-experimental Designs

Non-Experimental DesignsCorrelational studies, observational

studies, quasi-experimental studies, and surveys

Experimental (or True) DesignsManipulate the IV(e.g., Control vs. Experimental groups)Random assignmentConfounds are controlled or eliminated

Page 11: Research Methods: Design and Analysis

Research Designs

To schedule a consultation call 877-437-8622 or email [email protected]

Experimental Designs

One shot case studies X O

One group pretest posttest O1 X O2

Statistic group X O1 O2 R=random assignment, O=observation, X=treatment group

Page 12: Research Methods: Design and Analysis

Research Designs

To schedule a consultation call 877-437-8622 or email [email protected]

True Experimental DesignsPretest-posttest control group design

R O1 X O2

R O3 -- O4

R=random assignment, O=observation, X=treatment groupGroup Time 1 Time 2 Time3

Random Treatment Mean X Mean

Assignment Control Mean Mean

No diff Sig diff

Pretest Posttest0

5

10

ControlTreatment

Page 13: Research Methods: Design and Analysis

Research Designs

To schedule a consultation call 877-437-8622 or email [email protected]

For great research design references, see:

Cook and Campbell ‘79 Campbell and Stanley ‘63 Creswell ‘05 Leedy and Ormrod ‘10 SocialResearchMethods.net

Page 14: Research Methods: Design and Analysis

A word about Sampling Method and Sample Assignment

To schedule a consultation call 877-437-8622 or email [email protected]

Random sample: each individual in the population has an equal chance to become a participant of the sample.

Random assignment: each participant of the sample has an equal chance of being assigned to a group (treatment or control).

Page 15: Research Methods: Design and Analysis

Instrumentation

To schedule a consultation call 877-437-8622 or email [email protected]

These are the tools that measure the outcomes—typically variables in the tools are categorized into independent variables (IV’s) and dependent variables (DV’s).

IV’s (or treatment variables) are typically manipulated by researcher.

DV’s (also called outcome variables or criterion variables) are responses to IV, caused by IV, or predicted by IV.

The variables should be reliable and valid.

Don’t make your own.

See Google scholar, review articles, Statistics Solutions directory.

Page 16: Research Methods: Design and Analysis

Instrumentation

To schedule a consultation call 877-437-8622 or email [email protected]

Questionnaires Interviews Observations Rating scales Achievement tests Personality

inventories Historical reports or

data

Page 17: Research Methods: Design and Analysis

Method of Data Collection (Procedures)

To schedule a consultation call 877-437-8622 or email [email protected]

After describing the sample and the instruments, tell the reader how you are going to systematically and objectively administer these instruments to these participants.

ParticipantsMeasures

Page 18: Research Methods: Design and Analysis

Validity

To schedule a consultation call 877-437-8622 or email [email protected]

You measure what you say you are measuring.

(Validity is more important than reliability)

Page 19: Research Methods: Design and Analysis

Internal Validity

To schedule a consultation call 877-437-8622 or email [email protected]

Eliminate confounds or competing explanations of differences on DV (E.g., Differences on Self-efficacy by gender…but the woman were 35 y.o. and men were 75 y.o.)

Threats to internal validity• Selection bias—you think they represent the population (Safeguard tip: randomly sample and randomly assign)• Pre-post studies without a control group

o History/maturation—something else occurred between the assessments

• Drop-out/attrition • Reliability of measure• Low power• Order effect• Multiple tests of significance • John Henry Effect

Page 20: Research Methods: Design and Analysis

External Validity

To schedule a consultation call 877-437-8622 or email [email protected]

Def: The ability to generalize outcomes to the population

Threats• Population validity—ability to generalize to

population. (E.g., Kohlberg moral development—males value individual rights, where Gilligan found that females valued relationships).

• Ecological validity—ability to generalize from lab to everyday environments.

Page 21: Research Methods: Design and Analysis

Reliability

To schedule a consultation call 877-437-8622 or email [email protected]

Inter-rater reliability (kappa coefficient)

Test-retest reliability Internal consistency—

Cronbach’s alpha

Page 22: Research Methods: Design and Analysis

Data Analysis Plan(what to include)

To schedule a consultation call 877-437-8622 or email [email protected]

Reliability of scales: Cronbach’s alpha

Descriptive statistics: Means, standard deviations, frequency and percentages.

A data analysis plan for EACH research hypothesis.

• Inferential: ANOVA, MANOVA, regression, time-series, SEM, Chi-square, correlation

• Non-inferential: descriptives (e.g., M, SD, %)

Assumptions of analyses and justification for analyses

Page 23: Research Methods: Design and Analysis

Sample Size

To schedule a consultation call 877-437-8622 or email [email protected]

N = ?

1. Effect Size—importance of differences

2. Power—likelihood of finding differences (typically .80)

3. Alpha (typically .05)

Page 24: Research Methods: Design and Analysis

1-1 Consulting

To schedule a consultation call 877-437-8622 or email [email protected]

[email protected]

877-437-8622

Page 25: Research Methods: Design and Analysis

We’re interested… Webinar Feedback and

What You’d LikePlease provide feedback on the webinar and tell us what

else you’d like to learn about in upcoming webinars.

Send comments to: [email protected]

Page 26: Research Methods: Design and Analysis

Thank You for your Participation

and Attention!

Join us for our next webinar on Wednesday, November 20th at

8:30pm ETResults Chapter:

Conducting, Interpreting, and Writing