AI PLS

One workbench, three methods — pick what fits your question: full PLS-SEM (TAM template, drag-and-drop canvas, 5,000-resample bootstrapping, complete quality report), PROCESS-style mediation and moderation (Hayes Models 4 & 1), and multiple regression. Plus an optional AI interpretation of your results. Your data never leaves this device. Free, keyless.

Step 1 · Data

Load your dataset

A plain numeric CSV: one row per respondent, one column per indicator (survey item), with a header row. Non-numeric columns are ignored; rows with missing values on used indicators are dropped listwise. Nothing is uploaded — parsing happens in this tab.

…or paste CSV text
Step 2 · Model

Define constructs and paths

Each latent construct gets a name and its reflective indicators (Mode A). Click a column chip to assign it — or drag any chip onto another construct's card to move it there. An indicator may sit in more than one construct (that's how repeated-indicators higher-order models work). The structural model must be recursive — no loops.

Structural paths
Step 3 · Estimate

Run the model

Method notes: PLS-SEM uses reflective (Mode A) measurement with the Lohmöller algorithm; bootstrap uses percentile confidence intervals with construct-level sign alignment; p-values use the normal approximation to the t statistic. Results can differ slightly from SmartPLS 4 (initialization, sign-correction, and stopping details vary between implementations). Mediation and moderation run PROCESS-style Models 4 and 1 (Hayes) on observed scale scores — for other model numbers use dedicated software. Not included, honestly: covariance-based SEM (CB-SEM needs maximum-likelihood covariance fitting — use lavaan or AMOS and cite them), GSCA, and formative (Mode B) measurement within PLS. Independent tool, not affiliated with SmartPLS GmbH; all methods implemented from the published literature.

Scholarship

Methods & attribution

AI PLS is an independent implementation of methods from the published literature. Cite the original methodologists in your write-up — these are the canonical sources for what each part of your results rests on.

Method in this toolCanonical source
PLS path modeling algorithmWold (1982); Lohmöller (1989)
Bootstrap resampling & CIsEfron (1979)
Cronbach's αCronbach (1951)
Composite reliability, AVE, Fornell–Larcker criterionFornell & Larcker (1981)
ρA (consistent reliability)Dijkstra & Henseler (2015)
HTMT discriminant validityHenseler, Ringle & Sarstedt (2015)
f² effect-size thresholdsCohen (1988)
Mediation & moderation (Models 4 / 1), simple slopesHayes (2022); Aiken & West (1991)
TAM model templateDavis (1989)
PLS-SEM reporting practice (loadings ≥ .708, 5,000 resamples…)Hair, Hult, Ringle & Sarstedt (2022)