# Reproduce the Essai 1 calculations

The input is a literal transcription of Table 1 in the final publisher-formatted Oraee et al. (2017) article. It is not a new content analysis of the underlying 62 studies. Read [findings](findings.txt) for the question, result and limits.

Files:

- `table-1.csv`: 62 rows, five binary visible-mark columns, source locators and explicit blank-cell interpretation.
- `table-1.json`: the same matrix plus source hash, version and separate reported summaries.
- `analyse.py`: Python 3 standard-library calculation, reconciliation, all 1,953 single/pair omission cases and recoding thresholds.
- `results.json`: deterministic result of running the script.
- `transcription-check.json`: visual/extraction verification and a retained cold-review disagreement.
- `test_analyse.py`: repository-only checks; intentionally excluded from website payload.

In a repository checkout (tests are not included in website downloads):

```sh
python3 analyse.py
python3 -m unittest discover -s . -p 'test_*.py' -v
```

The script reads adjacent files and writes `results.json`. Use `python3 analyse.py --output /chosen/path/results.json` to retain an earlier output. It does not retrieve sources, send data or alter the matrix. All calculations can run offline. An unmarked cell means no visible dot, not an independently confirmed absence of a topic.

The source is linked through the [author's publication list](https://elenip.com/publication/) and the [DOI](https://doi.org/10.1016/j.ijproman.2017.07.001). The PDF is not redistributed here. `table-1.json` records the exact author-hosted copy and its hash.


# Independent cold visual verification — final Oraee Table 1

1 November 2026. Counted from rendered final publisher PDF without reading the reconstruction dataset. Poppler rendered PDF pages 9–10 (= journal pages 1296–1297) at 160 dpi. Renderings: final-09.png and final-10.png. PyMuPDF was unavailable in the bundled interpreter; Poppler used successfully.

## Observed column counts

| Category | Visible marks | Printed narrative count | Printed share |
|---|---:|---:|---:|
| Context | 20 | 22 (p1295 §4.2.1) | 35% |
| Process | 42 | 41 (p1297 §4.2.2) | 66% |
| Task | 8 | no count in §4.2.3; 13% | 13% |
| Team | 18 | 18 (p1297 §4.2.4) | 29% |
| Actor | 13 | 14 (p1297 §4.2.5) | 23% |

Total visible marks: 101 across 62 listed rows. Printed percentage row is at top of p1297 beneath row62, denominator stated as total studies focused on collaboration (62).

## Independent row-index check

Context: 1,4,10,11,13,15,16,24,32,33,34,35,40,44,50,53,54,57,58,60.
Process: 2,4,5,7,8,10,11,12,14,15,18,19,20,21,22,23,25,26,27,28,29,30,32,35,37,38,39,40,42,45,46,47,48,49,51,52,53,54,55,56,58,60.
Task: 5,13,32,37,47,49,59,62.
Team: 6,9,12,15,17,23,30,31,34,35,36,39,41,47,51,52,55,59.
Actor: 3,6,19,23,30,31,33,38,39,47,48,49,61.

Category-count breadth: 0=1 row; 1=30 rows; 2=23 rows; 3=7 rows; 4=1 row; 5=0 rows.

## Rows requiring interpretive caution

- Row43 Ochieng and Price (2010), p1296: no visible mark in ANY of five columns. It is named as a Context example in p1295 prose, but do not silently fill this cell. This is an unresolved printed-table gap, not proof the underlying research assigned no category.
- Row21 Gu and London (2010), p1296: Process mark only. Narrative p1297 §4.2.5 names it among Actor examples, but no Actor dot is visible in Table1. Do not silently add it.
- Row49 Pikas et al. (2013), p1296: Process/Task/Actor; no Context dot, although p1295 narrative uses it in discussion of BIM education and Context. This might be another reconciliation lead, not permission to change the transcription.
- All other visible dots were legible at the rendered resolution; no visually uncertain dot assignment remained.

No Figure8 was found by full-paper text scan in this final 14-page PDF. The final paper's figures run to Fig6. If a manuscript copy contains Fig8, do not treat its claims as part of this final version without an explicit version crosswalk.

## Consequence

The visible-table counts do not reproduce the final printed Context, Process and Actor totals. Process's rank lead nevertheless remains in either reading (42 vs20 visible; 41 vs22 reported). This supports an explicit two-baseline sensitivity analysis. It does NOT establish a cause (typesetting, omitted marks, earlier coding state or other explanation), misconduct, invalid qualitative coding, or a need for retraction. Author/editor clarification would be a separate authorised action.

## Cold-count correction retained

The first manual count reported Process41 because I missed its dot on row32 Kihong and Mahabaleshwarkar (2011). An independent pixel-column component count detected42 marks. I then inspected an enlarged crop (`final-row32-crop.png`) and confirmed row32 has Context + Process + Task. The corrected independent total is42 and total marks101. The reconstruction analyst independently identified the same row. This was an audit transcription miss, not a change to the PDF. Pixel-component counts for p1296 confirmed Context20, Process42, Task7, Team18, Actor13; row62 on p1297 adds the eighth Task.