Fold Excel and LaTeX into tool_notes, write the two qualitative-method pages, correct the family description

CLAUDE.md

@@ -227,9 +227,11 @@ unpublished; both were true once and neither is now.
 
 The six calculators are the largest design surface in the stack family and the
 obvious place to start. Beyond the stacks: Experiments, openstacks.dev and the
-ImpactMojo properties. SignalStack is dead and ViewStack, BridgeStack and
-RootStack are archived; the live family is InsightStack, FieldStack,
-EquityStack and PolicyStack under OpenStacks-for-Change.
+ImpactMojo properties. The family is three stacks, InsightStack, FieldStack and
+EquityStack, with openstacks.dev as the index. SignalStack is the Research
+Rundown companion archive, beside the stacks rather than one of them.
+PolicyStack is superseded by PolicyDhara; RootStack, BridgeStack and ViewStack
+are archived.
 
 One constraint that catches people: `Experiments` serves under a strict Content
 Security Policy allowlisting specific CDNs, so a design pulling fonts or scripts

README.md

@@ -44,8 +44,7 @@ This is the **knowledge systems layer** of [OpenStacks for Change](https://opens
 | Directory | What It Does | Tool |
 |-----------|-------------|------|
 | `vensim/` | 8 system dynamics models (health, agriculture, climate, migration, education) | Vensim |
-| `tool_notes/` | Seven visual and collaboration tools (Excalidraw, Kumu, Observable, RawGraphs, Flourish, Power BI, Miro): when to use each, when not, what goes wrong, one worked file apiece | Notes |
-| `excel_visuals/` | Excel visualisation templates | Excel |
+| `tool_notes/` | Nine tool notes (Excalidraw, Kumu, Observable, RawGraphs, Flourish, Power BI, Miro, Excel, LaTeX): when each is worth the setup, when it is not, what goes wrong, with one worked file each | Notes |
 
 ### Knowledge and Documentation
 
@@ -57,9 +56,8 @@ This is the **knowledge systems layer** of [OpenStacks for Change](https://opens
 | `KM_tools/` | Folder structure, project README template, file naming, tagging | Markdown |
 | `annotated_research/` | Five research briefs marked up with one eight-tag annotation scheme, weakest design to strongest; illustrative, for teaching critical reading | Markdown |
 | `learning_layers/` | Embedding learning in the MEL cycle as scheduled decisions at three tempos | Markdown, PDF |
-| `taguette_coding/` | Qualitative coding workflows |
-| `visual_ethnography_descript/` | Ethnographic documentation tools |
-| `latex/` | LaTeX templates for research reports |
+| `taguette_coding/` | One focus group coded start to finish in Taguette, with the method: frame first, double-code two transcripts, count carefully | Qualitative |
+| `visual_ethnography_descript/` | From recording to clip to quote: three consents, quotes checked against audio, clips cut with context visible | Qualitative |
 | `workflows/` | End-to-end analysis workflow guides |
 
 ## Getting Started
@@ -83,10 +81,11 @@ InsightStack is one of several stacks in the [OpenStacks](https://openstacks.dev
 
 | Stack | Focus |
 |-------|-------|
-| **InsightStack** (this repo) | MEL tools, calculators, documentation |
-| [FieldStack](https://github.com/Varnasr/FieldStack) | Field operations and survey analysis in R |
-| [EquityStack](https://github.com/Varnasr/EquityStack) | Distributional analysis and survey estimation in Python |
-| [PolicyStack](https://github.com/Varnasr/PolicyStack) | 15 flagship schemes, four years of budget data, performance indicators; superseded for new work by [PolicyDhara](https://github.com/Varnasr/PolicyDhara) |
+| **InsightStack** (this repo) | MEL tools, calculators, research documentation, and loaders for real survey microdata. Stata, Python, R, SPSS |
+| [FieldStack](https://github.com/Varnasr/FieldStack) | Field operations while a survey is being collected, and survey analysis after. R |
+| [EquityStack](https://github.com/Varnasr/EquityStack) | Equity from a development economics perspective: distributional analysis, the concentration index, and design-based survey estimation. Python |
+
+[openstacks.dev](https://openstacks.dev) is the index for all of it. [SignalStack](https://github.com/Varnasr/SignalStack) is the companion archive for the [Research Rundown](https://researchrundown.substack.com) newsletter, alongside the stacks rather than one of them. [PolicyStack](https://github.com/Varnasr/PolicyStack) is superseded by [PolicyDhara](https://github.com/Varnasr/PolicyDhara). RootStack, BridgeStack and ViewStack are archived.
 
 ## Contributing
 

index.html

@@ -157,20 +157,16 @@ <h2>Code and tooling</h2>
       <span class="d">Where an intervention pushes back on itself: eight system models across health, agriculture, climate, migration and education.</span>
     </a>
     <a class="row" href="https://github.com/Varnasr/InsightStack/tree/main/tool_notes">
-      <span class="t"><b>Tool notes</b><span>7 tools</span></span>
+      <span class="t"><b>Tool notes</b><span>9 tools</span></span>
       <span class="d">Excalidraw, Kumu, Observable, RawGraphs, Flourish, Power BI and Miro: when each is worth the setup, when it is not, and what goes wrong, with one worked file apiece.</span>
     </a>
-    <a class="row" href="https://github.com/Varnasr/InsightStack/tree/main/excel_visuals">
-      <span class="t"><b>Excel visuals</b><span>Excel</span></span>
-      <span class="d">Charts for the colleague who will only ever open Excel.</span>
-    </a>
     <a class="row" href="https://github.com/Varnasr/InsightStack/tree/main/taguette_coding">
-      <span class="t"><b>Taguette coding</b><span>Qualitative</span></span>
-      <span class="d">Code interview transcripts and get the tags back out for analysis.</span>
+      <span class="t"><b>Qualitative coding</b><span>Taguette</span></span>
+      <span class="d">One focus group coded start to finish, and the method: decide the frame first, code two transcripts twice, count carefully. Eleven of fourteen groups is a sentence; 78 per cent of respondents is a survey result a focus group cannot give.</span>
     </a>
     <a class="row" href="https://github.com/Varnasr/InsightStack/tree/main/visual_ethnography_descript">
-      <span class="t"><b>Visual ethnography</b><span>Descript</span></span>
-      <span class="d">A Descript storyboard and a climate fieldwork transcript, showing how a recording becomes notes.</span>
+      <span class="t"><b>Recorded fieldwork</b><span>Descript</span></span>
+      <span class="d">From recording to clip to quote, and what is decided at each step. Three separate consents, every quote checked against the audio, and a clip cut with its context visible to whoever approves it.</span>
     </a>
     <a class="row" href="https://github.com/Varnasr/InsightStack/tree/main/annotated_research">
       <span class="t"><b>Annotated research</b><span>5 briefs</span></span>
@@ -196,10 +192,6 @@ <h2>Code and tooling</h2>
       <span class="t"><b>Knowledge management</b><span>Conventions</span></span>
       <span class="d">One folder structure and a README template, so a file's location says what it is. Raw data is never edited; every table came from a script still in the folder.</span>
     </a>
-    <a class="row" href="https://github.com/Varnasr/InsightStack/tree/main/latex">
-      <span class="t"><b>LaTeX templates</b><span>LaTeX</span></span>
-      <span class="d">A LaTeX quick-start template, with notes on where it helps.</span>
-    </a>
   </div>
   <p class="note">These open on GitHub, where the code lives.</p>
 

taguette_coding/README.md

@@ -1,11 +1,58 @@
-# Taguette Coding
+# Qualitative coding with Taguette
 
-This folder contains a real field-based qualitative transcript coded using [Taguette](https://taguette.org), an open-source tool for visual tagging.
+One coded focus group discussion, start to finish: the transcript, the tag
+file Taguette produced, and the export that goes to analysis. This is the
+worked example for the qualitative half of a mixed-methods evaluation, and
+the method note that goes with it.
 
-## 📂 Folder Contents
-- `chhattisgarh_chw_fgd.txt` – Real translated field transcript from a community health worker
-- `chhattisgarh_chw_fgd.tag.json` – Taguette-compatible tag file with three theme codes
-- `tag_summary_export.html` – Visual export of tags with sample quotations
+| File | What it is |
+|---|---|
+| `chhattisgarh_chw_fgd.txt` | A focus group discussion with community health workers, as the transcript Taguette reads. Illustrative, not a real group |
+| `chhattisgarh_chw_fgd.tag.json` | The tags applied to it, with the highlighted spans, in Taguette's own export format |
+| `tag_summary_export.html` | Every highlight grouped by tag, which is the document you actually analyse from |
 
-## 🧠 Use Case
-Taguette supports thematic coding of qualitative transcripts like FGDs and interviews. This example is from a CHW in rural Chhattisgarh describing climate-driven fieldwork adaptations. Useful for researchers, MEL teams, and documentation practitioners.
+Taguette is free, open source, and runs locally (`pip install taguette`,
+then `taguette`) or on the hosted service. It does one thing: highlight a
+span of text and attach a tag. That is the whole of qualitative coding, and
+the expensive tools do not do it better.
+
+## The method, in the order it happens
+
+**1. Decide the coding frame before opening the transcript.** A codebook of
+eight to fifteen tags, each with a one-line definition and an example, drawn
+from the evaluation questions. Tags invented while reading are fine and
+should be added, but a frame that exists only in the coder's head cannot be
+applied by a second coder, and a second coder is the whole check.
+
+**2. Code the same two transcripts independently, then compare.** Where two
+coders tag the same passage differently, the definition is ambiguous. Fix
+the definition, not the coder. Do this before coding the other thirty.
+
+**3. Code for what is said, not for what you expected.** The tag
+`barrier_transport` goes on a passage about the bus not coming, whether or
+not the evaluation question was about transport. Tags for the questions you
+came with are necessary; tags for what the respondents brought are where the
+finding is.
+
+**4. Export by tag, and read each tag's passages together.** That is what
+`tag_summary_export.html` is. A theme is a tag whose passages, read
+together, say something the individual quotes did not.
+
+**5. Count, carefully.** "Eleven of fourteen groups raised transport
+unprompted" is a defensible sentence. "78 per cent of respondents cited
+transport as a barrier" is the same fact dressed as a survey result, and a
+focus group is not a sample.
+
+## What this cannot do
+
+Taguette has no query language, no co-occurrence matrix, no inter-coder
+statistic. For a study with a hundred transcripts and three coders, that
+matters, and NVivo or MAXQDA earn their price. For an evaluation with
+fifteen focus groups and one coder checked by a second, they do not.
+
+## Related
+
+`../visual_ethnography_descript/` for the same discipline applied to recorded
+rather than transcribed material. `../annotated_research/annotation_scheme.md`
+tag **[MEASURE]** for what a self-reported answer in a group setting is
+evidence of.

tool_notes/README.md

@@ -1,13 +1,13 @@
 # tool_notes
 
-Short, honest notes on seven visual and collaboration tools, with the one
+Short, honest notes on nine visual, document and collaboration tools, with the one
 worked file each note needs to make its point. When the tool is worth the
 setup, when it is not, and what will go wrong.
 
-These were seven top-level folders, each presented as a module. None of them
+These were nine top-level folders, each presented as a module. None of them
 is: the tools live on their own sites, and what a repository can usefully hold
 about them is a page of judgement plus a small example, which is what each
-folder now contains. The seven are here together so nobody mistakes a
+folder now contains. They are here together so nobody mistakes a
 screenshot for a template again.
 
 | Note | Tool | What is actually in the folder |
@@ -19,6 +19,8 @@ screenshot for a template again.
 | [`flourish_charts/`](flourish_charts/) | Flourish | The iframe embed snippet, and a screenshot of a chart |
 | [`powerbi_reports/`](powerbi_reports/) | Power BI | An adaptation note. No `.pbix` ships; the note links to Microsoft's sample |
 | [`miro/`](miro/) | Miro | A human-centred design session outline and links. No board ships |
+| [`excel_visuals/`](excel_visuals/) | Excel | Three chart templates as `.xlsx`: bullet, waterfall, Gantt |
+| [`latex/`](latex/) | LaTeX | A report skeleton that compiles with `pdflatex`, and where it is worth the learning curve |
 
 ## The one judgement that applies to all seven
 

tool_notes/excel_visuals/README.md

@@ -0,0 +1,36 @@
+# Excel
+
+Three chart templates as `.xlsx`, for the colleague who will only ever open
+Excel. Each has a data sheet you overwrite and a chart sheet that redraws.
+
+| File | What it makes | Use it for |
+|---|---|---|
+| `bullet_chart_template.xlsx` | A bar against a target line and shaded bands | Target versus actual, one indicator per row |
+| `waterfall_chart_template.xlsx` | Rising and falling bars from a start value to an end value | What moved a total between two periods |
+| `gantt_chart_template.xlsx` | Horizontal bars on a date axis | A field schedule, one row per activity |
+
+## When to use it
+
+The audience will edit the numbers themselves, and they work in Excel. A
+programme manager updating a bullet chart each quarter needs a file they can
+open, not a script they cannot run. These three cover the charts a monitoring
+report needs and a default Excel chart does badly.
+
+## When not to
+
+Anything that will be produced more than once from data that arrives as a
+file. Paste-into-a-template is a manual step every time, and it is the step
+that gets done wrong at 6 p.m. before a deadline. FieldStack's
+`custom_viz/` has the bullet and waterfall charts as R code that reads the
+data and draws.
+
+## What goes wrong
+
+Excel charts reference cell ranges. Adding a row below the data does not
+extend the chart; the new row is silently left off. Convert the data range to
+a Table (Ctrl+T) before pointing the chart at it, and it grows.
+
+Colours are set per series. A chart copied into a report with the
+organisation's palette will need each series recoloured; there is no theme
+file here, because Excel's chart themes do not travel reliably between
+machines.

tool_notes/latex/README.md

@@ -0,0 +1,41 @@
+# LaTeX
+
+A quick-start template for a report, and notes on where LaTeX is worth the
+learning curve for a development researcher.
+
+| File | What it is |
+|---|---|
+| `quick_start_template.tex` | A report skeleton: title, abstract, sections, a table, a figure, a bibliography stub. Compiles with `pdflatex` and nothing else installed |
+| `field_use_case_latex.md` | Where it has earned its place in this line of work |
+| `community_links_latex.txt` | Where to ask |
+
+## When to use it
+
+A document with many numbered equations, a document with fifty tables that
+have to look identical, or anything going to a journal that supplies a class
+file. And a working paper where the bibliography is a `.bib` file that will
+be reused: BibTeX is the reason to learn LaTeX, and it pays back on the
+second paper.
+
+## When not to
+
+A report co-written with a programme team. Word's track changes is how most
+of the sector reviews a document, and a LaTeX source is unreadable to a
+reviewer who does not write it. Overleaf softens this and does not remove it.
+For a donor report, a policy brief, or anything a non-technical colleague
+must edit, use Word or Quarto, which renders to Word.
+
+Quarto is the honest middle: Markdown with code chunks, rendering to PDF
+through LaTeX when you want that and to Word when you do not. FieldStack's
+`notebooks/` and `automated_reporting/` are Quarto.
+
+## What goes wrong
+
+Tables. A regression table pasted from Stata or R into LaTeX by hand is
+wrong by the third revision. Use `esttab` in Stata or `modelsummary` in R to
+write the `.tex` directly, and `\input` it, so the table in the paper is the
+table the code produced.
+
+Fonts for Indic scripts need `xelatex` or `lualatex` rather than `pdflatex`,
+and a font that has the script. `pdflatex` will compile a document with
+Hindi in it and print boxes.