From statistics basics to using statory

Text and discourse analysis — reading the frames in what is said

When to use it. Use it when your material is words already out in the world — news articles, speeches, policy papers, social media. If interviewing is "asking people," discourse analysis is "re-reading what was said."

What to look for. The same event becomes a different story depending on its frame. Frame analysis reads each text while marking four functions (Entman): (1) what it defines as the problem, (2) where it puts the cause, (3) what it calls right, and (4) what remedy it offers. Those four are statory's default codes, and you can rewrite them to fit your material.

Steps. Split into units (each holding one claim) → code the frames → cross them by speaker and outlet (who speaks in which frame) → contrast the wording (which words each frame prefers). The crosstab and the wording contrast are where qualitative work meets quantitative — frame counts can be tested with a chi-square crosstab.

Writing it up. Your methods section should give the definition of the unit of analysis, where the coding frame came from (Entman applied deductively, or induced from the material), and coding reliability (kappa, if two people coded). statory's audit-trail summary gives you the material for that section.

Phenomenology — describing the essence of an experience

When to use it. When your research question is "what is it like to go through this?" The subject is the meaning of lived experience that will not reduce to a number — a day in someone's recovery, the shock of losing a job for the first time. The material is a small set of in-depth interviews, usually 5 to 10 people.

The core stance — epoché (bracketing). Writing down what you already believe, before you analyze, is part of the method. You cannot erase your assumptions, but once they are on paper you can catch the moment you start reading the material through them. In statory you leave them as a prior-understanding memo when the analysis begins.

Steps (based on Giorgi's four). Read the whole thing (all the way through, marking nothing) → mark meaning units (wherever the meaning turns) → transform the meanings (put the participant's words into scholarly language — the meaning stays theirs) → describe the essential structure (write, in your own sentences, the structure that runs through those meanings). Colaizzi's approach differs mainly in where the theme clusters are gathered, and statory accepts both flows.

About AI. statory suggests candidate meaning units and stops there; there is no AI button for transforming meanings or describing the essence. In this method those two steps are the researcher's interpretive work itself — when a reviewer asks "how did you arrive at this essential description?", "the AI wrote it" is not an answer you can defend.

Writing it up. Reviewers judge the depth of the description, not the number of participants. Build the results section from the essential structure plus one representative quote per meaning unit; statory's quote lookup by code helps you choose them.

Narrative analysis — reading a life as a story

When to use it. When the path of one life (or a few) is itself the subject — life histories, recovery accounts, autobiographical records. The unit is a person, not a theme.

The core idea — time and turning points. The order someone tells it in is not the order they lived it. Put the pieces of the story back on a timeline and mark the turning points where the direction of the life bent, and the before and after start to explain each other. That is what statory's timeline rebuild and turning-point marking do.

Steps (Riessman's thematic analysis). Split into segments (one scene, one event) → rebuild the timeline → mark the turning points → code the themes (the threads that run across the story) → describe the story's shape (beginning, unfolding, turn, now). To go as far as structural analysis (Labov), you can also mark each segment's function: abstract, orientation, complication, resolution, evaluation.

Writing it up. A narrative results section usually rebuilds "the case's story" with quotes in time order. statory arranges those quotes chronologically as your first draft. Anonymizing is not optional — use pseudonyms, handle identifying details, and always pass the qualitative anonymization gate.

Grounded theory — drawing a theory out of the material

When to use it. When you are building the frame of a theory for something no existing theory explains, or explains badly. The question takes this shape: what brings this about, how do people respond, and what is left afterward?

Steps (Strauss & Corbin). (1) Open coding — name every concept (name a lot, tidy up later). (2) Build categories — turn codes into categories by constant comparison (group, ungroup, group again). (3) Axial coding — place the categories in the paradigm: causal conditions → the central phenomenon ← context and intervening conditions, action/interaction → consequences. (4) Selective coding — the core category and the storyline. (5) Theoretical saturation — until new material stops producing new codes. statory's saturation curve is a reference, not a verdict; you make the call — and in the paper write "the grounds on which the researcher judged saturation," not "the curve decided."

The axial coding diagram. Your paradigm layout is the axial coding figure in the paper. In statory the layout is the source and the diagram is a view of it, so the figure and the analysis cannot drift apart.

From qualitative to quantitative — extending this method. Categories built with grounded theory can become the raw material for survey items (an exploratory sequential mixed design). The order is strict: categories → draft items → collect data → check reliability (alpha) and factor structure (EFA, CFA) → structural equation modeling. SEM that skipped validation will not survive review, and statory does not provide a button that skips these steps.

Writing it up. Three questions come up again and again about the methods section: on what grounds you judged saturation, how you secured coding reliability (memos, audit trail, two coders), and how the axial model was derived. statory's process-summary export gives you evidence for all three.

Consensual qualitative research (CQR) — reading together, concluding by consensus

When to use it. When you have a research team (usually 3 to 5 people) and want the trustworthiness of the conclusion to rest on the team's consensus rather than one person's reading. The typical material is semi-structured interviews about concrete experience, roughly 8 to 15 cases.

The core structure — three layers of checking (Hill). (1) Each team member codes without seeing the others. (2) Where they read it differently, they trade reasons and reach consensus. (3) An auditor who took no part in the analysis reviews the result. The spirit of the method is that this is an exchange of reasons rather than a vote, and that the team even decides whether to accept the auditor's points — but records the decision either way.

Steps. Code domains and core ideas → reach consensus → auditor review → cross-analysis. In the cross-analysis you count how many cases each category appears in and label it General (nearly every case), Typical (more than half), or Variant (fewer). statory works these labels out for you, but with a small sample it will not quietly move the cutoffs — it asks you to state in the paper which cutoffs you used, because studies stop being comparable once every paper picks its own.

In statory. Each member's coding appears in the double-coding comparison (with kappa shown alongside), consensus happens in the side-by-side view, and auditor comments — accepted or rejected — stay in the audit trail. That whole record is your answer when a reviewer asks how the consensus process was secured. A finished codebook can be saved to the library and applied deductively to your next material — but loading it clears the code assignments. You inherit the categories; the judgments have to be made again on the new material.

Writing it up. The standard results table is category by frequency (G/T/V) plus a representative quote for each category. statory's three paper-ready table exports — theme structure, quotes, and CQR frequencies — give you the skeleton of that section as they are.

Sociometry — drawing relationships as a map

When to use it. When the question is the structure of relationships itself inside a bounded group — a class, a team, an organization, a community — who chooses, leans on, or pushes away whom. The material comes in two kinds: surveys (nominations, such as "name three people you would like to work with," or ratings, where every pair of members is rated on a Likert scale) and text (relational statements inside interviews and field notes).

Starting from text. This path is particular to statory. Statements in an interview such as "A leans on B" are pulled out as (who → whom, in what relation, how strongly) — AI brings the candidates, but only where the supporting sentence really exists — and once you confirm them they become a matrix. It is the bridge that carries relationships you found qualitatively into a quantitative network, and it ends the back-and-forth between NVivo and UCINET in one place.

Why ratings are worth it. Nominations leave you with chosen or not chosen; ratings catch the asymmetry — A gives B a 5 while B gives A a 2. Most of a group's tension lives in that asymmetry.

How to read it — four structures. On the map, look for the central figures (where choices pile up), members no connection reaches, small groups that are close only to each other (cliques), and the people who bridge separate clusters. Status and rejection indices, cohesion, and density put numbers under all four.

Ethics — the weight of this method. Asking who rejects whom can leave a mark on a group. Use negative nominations only when you truly need them, and only with anonymity assured. When you share results, the rule is not to show a group a sociogram in which its own members can be identified.

Writing it up. The methods section states the collection technique (fixed-choice nominations, free nominations, or ratings) and how you drew the boundary (who counts as part of the group); the results section carries the sociogram plus a table of indices. If you measured at two or more time points, comparing the change in density and cohesion becomes evidence for the effect of an intervention.