Loading…
VibeLux's regression engine fits a model on your own facility's logged data — correlating your light, climate, and nutrient drivers against the yield and quality outcomes you measure. It surfaces which drivers track with an outcome in your data; it is correlational and advisory, not a trained predictive model and not a guaranteed result.
An ordinary-least-squares model fitted on the data you log — with the diagnostics to tell you how much to trust it
A feature-builder pulls your logged light, climate, and nutrient drivers over the window that matches each outcome — no external corpus, just your facility's history.
Coefficients with p-values, R²/adjusted-R², VIF collinearity checks, partial correlations, and leave-one-out cross-validation — so you see how strong the relationship really is.
Below ~6 samples it returns an honest "not enough data yet" instead of fabricating a trend. More samples make the fit more trustworthy.
Enter yield/quality results — COAs, weights, lab values
Pull your light/climate/nutrient drivers for each window
Least-squares with VIF & partial-correlation checks
Leave-one-out CV + residual diagnostics
See which drivers track the outcome — advisory
The same engine works for any measurable outcome you log — across cannabis, produce, biomass, and bioactive compounds
Correlate spectrum and climate against your logged cannabinoid and terpene COAs
Example outcomes per sector, with the size of each addressable market for context
Where we'd like to take outcome analysis next — not yet shipping
Treatment-vs-control experiment design with interaction analysis, beyond single-factor correlation
More quality analytes — carotenoids, polyphenols, firmness, shelf-life — with per-crop templates
Attribute outcomes back to the exact spectral recipe and trial that produced them
Log what you measure, and let the regression show you which light, climate, and nutrient drivers actually track with it — honestly, on your own data.
Runs on your own data • Honest fit statistics • Correlational & advisory