Regression & Advanced¶

Linear Regression¶

Performs ordinary least-squares (OLS) linear or polynomial regression.

Details

Set Degree > 1 for polynomial regression (e.g. 2 = quadratic, 3 = cubic). With degree 1 (default), this is standard linear regression.

Outputs:

coefficients — slope, intercept, standard error, 95% CI, and p-values per parameter
residuals — fitted values, residuals, and standardized residuals for downstream plotting

Summary statistics: R², adjusted R², F-statistic, and F p-value.

      R-squared, coefficient, residuals, predict, fitted values,
      multiple regression, quadratic, cubic, standard curve, Bradford,
      線性回歸, 多項式迴歸, 迴歸分析, 最小二乘法, 斜率, 截距, 決定係數

Direction	Port	Type
Input	`in`	table
Output	`coefficients`	stat
Output	`residuals`	table
Output	`curve`	table

Properties: Polynomial Degree, ``

Nonlinear Regression¶

Fits nonlinear curves to XY data using scipy.optimize.curve_fit.

Details

Built-in models:

4PL (EC50 / Dose-Response) — four-parameter logistic for IC50/EC50
Hill Equation — sigmoidal binding/dose-response
One-Phase Exponential Decay — single-rate decay to plateau
Two-Phase Exponential Decay — fast + slow decay components
Exponential Growth — unbounded exponential increase
Michaelis-Menten — enzyme kinetics saturation curve
Gompertz Growth — asymmetric sigmoidal growth
Sigmoidal (Logistic) — symmetric S-curve

Outputs best-fit parameters with 95% CI and a smooth predicted curve table.

Direction	Port	Type
Input	`in`	table
Output	`parameters`	stat
Output	`curve`	table

Properties: Model, X Min (0=auto), X Max (0=auto)

Model Predict¶

Predicts Y values from a fitted model and a new data table.

Details

Connect the model output from Linear Regression or Nonlinear Regression, then provide a table with the X column to predict on.

The node auto-detects the X column name from the model metadata. Override with the X Column field if the new table uses a different name.

Outputs the input table with an added Predicted column.

      Bradford, ELISA, 預測, 插值, 標準曲線

Direction	Port	Type
Input	`data`	table
Output	`out`	table

Properties: `,Inverse X Min (0=auto),Inverse X Max (0=auto)`

Two-Way ANOVA¶

Performs two-way analysis of variance with interaction term (Type II SS).

Details

Input must be in long format with two factor columns and one numeric value column.

Outputs:

anova_table — sum of squares, df, F-statistic, and p-value per source
group_means — mean, SD, SEM, and N for every factor combination

Direction	Port	Type
Input	`in`	table
Output	`anova_table`	stat
Output	`group_means`	table

Contingency Analysis¶

Tests categorical association using chi-square and Fisher's exact tests.

Details

Input types:

Raw Data (two columns) — a crosstab is built automatically from two categorical columns
Contingency Matrix — a pre-built matrix of observed counts

Outputs:

test_results — Pearson chi-square, Yates-corrected chi-square, and Fisher's exact (2x2)
observed_counts — the observed contingency table
expected_counts — expected counts under the null hypothesis

Direction	Port	Type
Input	`in`	table
Output	`test_results`	stat
Output	`observed_counts`	table
Output	`expected_counts`	table

Properties: Input Type

Survival Analysis¶

Performs Kaplan-Meier survival analysis with log-rank test.

Details

Input columns:

Time Column — duration or follow-up time
Event Column — 1 = event occurred, 0 = censored
Group Column (optional) — categorical grouping for multi-group comparison

Outputs:

km_table — survival function with 95% CI (feed into SurvivalPlotNode)
log_rank — omnibus log-rank test statistic and p-value
pairwise_stat — pairwise log-rank results with optional p-value adjustment
P-Adj Method — multiple comparison correction for pairwise tests.

Direction	Port	Type
Input	`in`	table
Output	`km_table`	table
Output	`log_rank`	stat
Output	`pairwise_stat`	table

Properties: P-Adj Method (Pairwise)

PCA¶

Performs principal component analysis (PCA) for multivariate data exploration.

Details

Outputs:

transformed — PC coordinates per sample (connect to ScatterPlotNode for PC1 vs PC2)
loadings — feature contributions per principal component
variance — eigenvalues and cumulative variance explained per component
Standardize — when enabled, applies Z-score normalization before decomposition.

Direction	Port	Type
Input	`in`	table
Output	`transformed`	table
Output	`loadings`	table
Output	`variance`	stat

Properties: ``

Mixed Effects Model¶

Fits a linear mixed-effects model (LMM) for hierarchical / nested data.

Details

Mixed-effects models are essential when observations are grouped (e.g. cells within wells, animals within treatment groups, repeated measures per subject). They estimate fixed effects (population-level trends) and random effects (group-level deviations) simultaneously.

Configuration:

y_col — dependent (response) variable.
fixed_cols — fixed-effect predictor(s), comma-separated.
group_col — grouping variable for random intercepts (required).
random_slope_col — optional predictor for random slopes.
REML — use Restricted ML (default) or Full ML estimation.

Outputs:

fixed_effects — coefficient table with SE, z-value, p-value, 95% CI.
random_effects — per-group random intercept (and slope) estimates.
summary — model-level statistics: log-likelihood, AIC, BIC, number of groups, ICC.

Direction	Port	Type
Input	`in`	table
Output	`fixed_effects`	stat
Output	`random_effects`	table
Output	`summary`	stat

Properties: ``