%202.png)
Data Analysis Chapter Development
Builds a results chapter organised by your research questions, reported completely, evidenced properly, and stopping before interpretation begins.
Free. Works on Claude, ChatGPT, Gemini or any assistant that accepts a skill file.
What this skill does
The method, encoded.
Results chapters fail in a few predictable ways. Interpretation leaks in, so the reader cannot tell what was observed from what is being argued, and the discussion chapter then repeats it. Statistics are reported as a bare p value, with no test statistic, no effect size, no confidence interval and no n, which makes the result impossible to evaluate and impossible to defend. Everything the software produced is included, so the answer to each research question is buried among tables that answer nothing. Themes are asserted with one short fragment each, stripped of the context a reader would need to judge the interpretation. And the analyses that produced nothing quietly disappear.
This skill fixes each. It builds the chapter's structure from the research questions, which solves the organisation and the selection problem at once. It enforces the results-discussion boundary at sentence level, where interpretation actually enters. It sets a complete reporting standard for every statistical result and applies it to null results too, since an effect size and an interval are what tell a reader whether you found no effect or simply could not detect one. It rebuilds themes with definitions, distribution and extracts long enough to be assessed, including the case that fits least comfortably.
Best used for
- Structuring a results chapter around research questions
- Reporting statistics to examinable completeness
- Presenting qualitative themes with judgeable evidence
- Separating results from discussion after they have merged
- Reporting null and unexpected findings honestly
- Deciding what belongs in the chapter and what in an appendix
Typical inputs
What you give it.
Actual analysis output from the actual data, Final research questions or hypotheses, The analysis plan from the methodology chapter, Coded transcripts, code frame and audit trail, Reporting constraints from the ethics approval, Departmental style guide and a recent accepted dissertation
Typical outputs
What you get back.
A chapter structure built from the research questions, A sample and data obtained section with response, exclusions and analysable n, An assumption checks and data conditions section reporting violations and remedies, Complete statistical reporting with statistic, df, exact p, effect size, interval and n, Null and unexpected results reported at the same completeness, Themes with definitions, distribution and attributed contextualised extracts, A stated three-way selection rule for chapter, appendix and neither, Self-contained numbered tables and figures, Joint displays for mixed designs with convergence characterised, A traceability and completeness audit
Method coverage
What the skill works through.
- Why results chapters lose marks
- Building the structure from your research questions
- The line between results and discussion, enforced sentence by sentence
- Establishing who and what the results describe
- Reporting assumption checks and missing data
- What a complete statistical result contains
- Reporting the analyses that produced nothing
- Presenting themes with evidence a reader can judge
- The volume problem and a stated selection rule
- Tables and figures that stand alone
- Reporting mixed designs at the integration point
- The traceability and completeness audit
Download
Free skill. One file.
Enter your email once. Every skill you download after that takes a single click.
How to install
Add the skill file and the five kernel protocols to a Claude Project, a ChatGPT Project, a Gemini Gem, or paste them at the top of any assistant conversation. Then give it your real research material, not a description of it.
Download skillQuestions
Common questions.
What should a results chapter include and what should go in the discussion?
Results carry the finding, its evidential support, its magnitude and direction, the base it rests on, and the descriptive pattern. The discussion carries comparison with prior literature, explanation of why a result occurred, mechanism, implications and evaluation. The practical test is to scan for "because", "suggests", "consistent with" and "this may be due to": each occurrence is a crossing that belongs in the next chapter.
How do I report a statistical test properly?
Every inferential result needs seven elements: the test name, the test statistic with its symbol, the degrees of freedom, the exact p value, an effect size with its name, a confidence interval where the statistic supports one, and the n. Then state the direction and magnitude in substantive terms. The effect size is not optional: with a large sample almost anything reaches significance, and with a small one a real effect may not.
Should I report a result that was not statistically significant?
Yes, in the same detail as a significant one, including the effect size and the confidence interval. Those are what tell a reader whether you found no effect or failed to detect one, which are different findings. Write it as "no statistically significant difference was found" with the full statistics, not as "there was no difference", which overstates it. Dropping a planned analysis because it produced nothing is selective reporting.
How much of my output should go in the chapter?
Apply a three-way rule and state it. In the chapter: anything that answers a research question, anything needed to judge such a result, and anything that contradicts the study's overall pattern. In an appendix: full model output, complete cross-tabulations, the code frame, and anything a sceptical reader might want to inspect. Nowhere: output that answers no question. If you cannot say which question a result answers, it does not belong.
How many quotes should I use per theme?
At least two, from different participants, long enough to carry their context, each attributed with a participant identifier. Select for the range within the theme rather than only its clearest expression, and include the participant who fits it least comfortably. A theme evidenced only by its best example is a theme nobody can evaluate.
Can I report percentages for qualitative findings?
Counts of participants are usually better than percentages, because a percentage on a base of nine implies a precision the method does not have. Some interpretive traditions reject prevalence claims altogether and use language of typicality and salience instead. Whichever convention your tradition follows, state which you are using.
What makes a table self-contained?
A reader who opens the page cold should understand it without the surrounding text: a number and a caption saying what is shown and in which population, the n, variables labelled in words rather than dataset names, units stated, the base for every percentage, and notes defining any abbreviation or symbol. Every table must also be cited in the text, and the text should state the finding rather than narrating the table in full.
The skill chain
Works well with.
Research where people already are.
Analyse it where you already work.
Yazi helps researchers conduct surveys, AI interviews and longitudinal research directly through WhatsApp.
