Practice answers: ai-dependence-serial-mediation

Original synthetic teaching data, CC0. Not the journal dataset.
Read https://quickpls.com/research/papers/ai-dependence-serial-mediation

Install R with cSEM 0.7.1, digest and jsonlite; run:
Rscript reference-calculations.R . NEW-OUTPUT-FOLDER

Use a new output directory. The script repeats the reference, not the QuickPLS app. See checks.json for the scope of the separate validation. profiles.json records the exact model or set choices. These are readable setup instructions, not an importable app project. Bootstrap indices are one-based; matching seed labels across programs need not produce matching rows. Keep every returned alternative and failed calculation.

REPEAT INDIRECT EFFECTS
After repeating the R fits, use Python 3 (standard library only):
python derive-indirect.py NEW-OUTPUT-FOLDER/ai-dependence-serial-mediation/reference.json NEW-OUTPUT-FOLDER/ai-dependence-serial-mediation/bootstrap-reference.csv NEW-INDIRECT-ANSWERS.json
This multiplies every route within each saved row-refit and takes the percentile endpoints. It also works with the retained reference-answer.json and bootstrap-reference.csv in this pack.
