1/71
Looks like no tags are added yet.
Analysis plan
Rules written before examining results that explain how you will organize and interpret data.
Data collection versus data analysis
Collection means gathering information. Analysis means sorting, comparing, and interpreting it to answer your question.
Qualitative, quantitative, and mixed methods
Qualitative = words and meanings. Quantitative = numbers and measurements. Mixed methods = both, with a plan for connecting the findings.
Four questions an analysis plan answers
What counts? How will I sort it? How will I check myself? How does it answer my research question?
Sample
The people, cases, or sources you actually collect data from.
Variable
Something that can differ or change, such as running experience or workout ratings.
Unit of analysis
The thing you analyze and draw conclusions about, such as a person, post, or school.
Reliability versus validity
Reliability = consistency when repeated. Validity = whether you measure what you intend to measure.
Code, theme, and codebook
Code = short label for data. Theme = broader pattern across data. Codebook = written code definitions, examples, and non-examples.
Peer pressure versus time pressure
Peer pressure = acting because of what others might think. Time pressure = acting because time is limited.
Qualitative analysis sequence
Prepare and read data, build a codebook, code consistently, group codes into themes, check contradictory cases, then answer the research question.
Thematic analysis
Find patterns across interviews, discussions, or open-ended responses by coding and grouping themes.
Content analysis with counting
Define categories and count their appearances in texts.
Comparative or case analysis
Compare cases using the same features in a matrix.
Narrative analysis
Examine stories for setting, turning points, outcome, and meaning.
Document or source analysis
Ask who created a source, when, why, and for whom.
Sentiment analysis
Classify text as positive, negative, or neutral using a tool or rubric.
Grounded-theory-style coding
Code line by line, constantly compare data, build categories, and develop an explanation from the data.
Intercoder versus intracoder reliability
Intercoder = agreement between different coders. Intracoder = consistency when the same person codes again later.
Checking coding agreement
Independently code the same sample, then calculate matches divided by total decisions times 100.
Mean, median, and mode
Mean = sum divided by count. Median = middle ordered value. Mode = most common value.
Range and standard deviation
Range = highest minus lowest. Standard deviation = spread around the mean.
Difference between two group averages: tool
A suitable t-test.
Difference among three or more group averages: tool
ANOVA.
Patterns or relationships in categorical counts: tool
Chi-square.
Two numerical measures rise and fall together: tool
Correlation, such as Pearson's r for a linear relationship.
One measure predicts another: tool
Simple regression.
Changed over time: tool
Trend analysis.
Pearson's r
Ranges from -1 to +1; shows direction and strength of a linear relationship.
Correlation versus causation
Correlation means variables are associated; causation means one produces a change in another.
p-value
Probability of a result at least as extreme as the one observed, assuming the null hypothesis is true.
p below .05 versus a large p-value
Below .05 = statistically significant; a large p-value does not prove there is no difference.
Statistical significance versus practical importance
A statistically significant difference can be too small to matter in practice.
Quantitative rules to set in advance
Inclusion and exclusion criteria, handling outliers and missing data, significance level, statistical tests, and sample-size plans.
Likert-scale items
Ordered ratings such as 1-5; equal gaps are not guaranteed.
Small samples
Can give less precise results and make differences harder to detect.
Action research versus evaluation research
Action research often compares before-and-after changes and reflections.
Evaluation research
Compares outcomes with success criteria set in advance.
Ethnography
Analyze field notes and observations for cultural patterns.
Experimental or quasi-experimental analysis
Compare treatment and comparison groups, often with t-tests or ANOVA.
Focus groups
Thematic analysis, watching for dominant speakers.
Observational studies
Consistent checklists and frequency counts, watching for selective recording.
Needs assessment
Compare what exists with what a community needs, then rank needs by frequency and seriousness.
Phenomenography
Identify different ways people understand or experience something and describe relationships between those categories.
Survey analysis
Closed answers = counts, percentages, and suitable group comparisons; open answers = qualitative coding.
Mixed-methods integration
Use interviews to explain survey findings, compare the two kinds of evidence, or use interview themes to build survey questions.
Analysis log
Dated record of decisions.
Reflexivity memo
Notes about your assumptions, experiences, and expectations.
Triangulation
Check a finding using another source or method.
Blind coding
Hides group labels or details that could influence judgments.
Independent review
An independent coder or observer checks a sample without seeing your answers.
Negative cases
Look for evidence that challenges your interpretation.
Confirmation bias
Focusing on evidence supporting your expectations.
Selection bias
Choosing participants or sources in ways that favor a conclusion.
Observer bias
Recording what you expect to see.
Sampling bias
Your sample misrepresents the target population or misses important groups.
Interpretation bias
Explaining results the way you hoped.
Publication bias
Positive or expected findings are more likely to be reported or highlighted.
Funding bias
A sponsor influences conclusions.
Anchoring bias
Giving too much weight to first impressions.
Cultural bias
Interpreting others only through your own cultural expectations.
Measurement bias
An instrument or scoring rule systematically distorts results.
Social desirability bias
Giving answers that make you look good.
Recall bias
Remembering past events inaccurately.
Acquiescence bias
Agreeing regardless of statement content.
Nonresponse bias
People who respond differ from those who do not.
Hawthorne effect
Changing behavior because you know you are being observed.
Demand characteristics
Guessing the researcher's expectations and responding accordingly.
Self-selection bias
Volunteers differ from the broader population.
Extreme response versus central tendency bias
Extreme response = choosing scale endpoints regardless of content.
Recall heuristic bias
Estimating frequency from examples that easily come to mind.
Strong analysis and honest conclusions
Choose analysis that fits the question, explain repeatable steps, report all findings, and connect limitations to your decisions.