Dataset: (none)
Info
Context help
Minimum grid size for the data sheet. Used on startup, recovery after errors, and new datasets.
Select columns (1 column = 1 variable). Default: all columns.
Compute a statistic horizontally across selected columns for each row (e.g. variability between repeated measurements).
Results are written to new columns immediately after the rightmost selected column.
Replace text within cells in selected columns (like Ctrl+H). All occurrences in each cell are replaced.
Build a repeating sequence for experiment layout (e.g. 1–10, each value 3×, whole block 2×).
Preview: …
Higher-order models use centered X internally (X − mean) for numerical stability; coefficients in results are shown in original scale.
Success event is encoded as the higher numeric value; failure as the lower value.
X = number of successes, N = total trials; require 0 ≤ X ≤ N.
Column headers above the sheet name the matrix columns; cells contain counts. Rows A,B,C,D or C1,C2,… are skipped automatically.
Expected proportions per category (normalized to sum 1; default = uniform).
Full HTML source of the current results (with embedded CSS), ready to paste or save.
Visual HTML will be saved to XML_output/<name>.xml
Click a column or function to insert it into the formula.
Column references: [Name] or C1, C2, …
Strictly positive numeric values only. Uses the same Box-Cox engine as DOE response surface analysis.
Each selected column is scaled independently. Empty cells stay empty.
Maps data toward normality using Johnson SL, SB, or SU (percentile fit, Shapiro p for selection).
Each selected column becomes a group in the output DB table (Group + Value).
Splits one DB table into separate columns — one column per group.
Columns
Preview
Alternative hypothesis
Data must be in DB format: one row per observation with columns for Factor A, Factor B, and the numeric response variable.
DB format: one row per observation. Select one numeric response and one or more predictors (Factor / Covariate). Hierarchical model with Type I (sequential) ANOVA.
Select at least 2 columns (each column = one factor level).
Each row: factor level in one column, numeric value in another.
Wide format: one row = one subject. Column 1 may hold Subject ID (text); select only the measurement columns (e.g. T0, T30, T60). Each subject must have a value in every selected column.
Example: | Subject | T0 | T30 | T60 | — check T0, T30, T60 only (not Subject).
DB format: inner factor is nested in outer (e.g. Machine nested in Factory). Each factory must have its own inner levels.
Select one or more columns.
Display mode
Select one or more columns to compare.
Each observation as a dot (steelblue, jittered). Red line = group mean. Hover a point for value and spreadsheet row.
Layout
Options
Plot values against a time or date column (ISO, dd.mm.yyyy, Excel serial, or numeric index).
Value column(s) (Y):
Pairwise scatter matrix for 2–6 numeric columns. Diagonal shows a mini histogram; off-diagonal cells show scatter plots.
Rank categories by frequency or summed counts. Bars are sorted descending; optional cumulative % line highlights the 80% rule.
Fit a quadratic response surface for two factors and plot predicted response as a contour map. Uses the same model as DOE Response Surface Analysis.
Select one or more columns. X-axis shows observation number.
Select at least two columns (groups) to compare.
Full 2k design — all factor combinations. Response column is left empty.
Screening design for many factors in few runs (±1 levels).
Fixed orthogonal arrays (Minitab catalogue) — supports 2-, 3- and mixed-level factors in one design.
Select factor columns (2-level coded −1/+1 or low/high) and response column. Center points (0) are optional.
Factor columns:
CCD — factorial core (−1/+1), axial star points (±α), center points. Extend an existing factorial or start fresh.
Fit a quadratic model (linear + squared + interaction terms) for 2–3 continuous factors.
Factor columns (select 2–3):
Set mean (μ) and standard deviation (σ) for each column below.
Set lower and upper bounds for each column (values in [Min, Max)).
Set location (μ) and scale (σ) of ln(X) for each column (values are always > 0).
Set scale (β, mean) for each column (values are always > 0).
Set shape (k) and scale (λ) for each column (values are always > 0).
Set shape (α) and scale (θ) for each column (mean = αθ, values > 0).
Set shape parameters α and β for each column (values in (0, 1), mean = α/(α+β)).
Set number of trials (n) and success probability (p) for each column (integers 0…n).
Set mean (λ) for each column (non-negative integers; variance = λ).
Set required successes (r) and success probability (p) for each column (failures before rth success, integers ≥ 0).
Set degrees of freedom (df) for each column (mean = df, values > 0).
Set numerator df1 and denominator df2 for each column (values > 0).
Set degrees of freedom (df) for each column (symmetric, mean = 0 for df > 1).
Set location (μ) and scale (s) for each column (symmetric, heavier tails than normal).
Select at least two numeric columns. Pearson r (linear), Spearman ρ (monotonic), optional covariance matrix.
Test for a single extreme value in a numeric column. Grubbs (normal assumption) and Dixon Q (small n).
Select a column to test for normality.
Select at least two treatment columns. Each row is one block (subject).
Select at least two columns (groups) to compare.
Select at least two columns (groups) to compare.
Select at least two columns (groups) to compare.
Select at least two columns (groups) to compare.
Select one or more measurement columns. For each characteristic enter LSL and/or USL (one-sided specs supported). Requires at least two numeric values per column.
I–MR — one measurement per row, single column. X̄–R / X̄–s / Me–R — one row per subgroup; each numeric column is one sample (n = number of columns). Nelson run rules apply to the primary chart (I or X̄); rule 1 also on the range chart.
Each numeric column is one sample; each row is one subgroup.
Only complete subgroups are used; any trailing rows with fewer than n values are ignored.
One row per subgroup (time period / batch). p and u support variable sample size — control limits change per row (step UCL/LCL on the chart). np and c assume constant inspection size / area.
Balanced study: each Part × Operator cell has the same number of replicates. One row per measurement. Standard AIAG coefficients K1, K2, K3 (2–5 reps, 2–5 operators, 2–15 parts).
% P/T = 100 × k × σ / (USL − LSL). Used when LSL and USL are set.
Balanced crossed study: each Part × Operator cell has the same number of replicates. One row per measurement. Variance components from two-way ANOVA (Part, Operator, Part × Operator).
% P/T = 100 × k × σ / (USL − LSL). Used when LSL and USL are set.
Destructive / non-replicable study: each operator measures unique samples from homogeneous batches. Part is nested in Operator (not crossed). Balanced design — same number of samples per operator, same replicates per sample.
% P/T = 100 × k × σ / (USL − LSL). Nested ANOVA variance components.
Crossed gage study with optional extra factors (lab, method, shift, …). Operator and Part are required; add more predictors below. Type I (sequential) GLM ANOVA — operator-related terms contribute to Reprod, part-only terms to Part-to-Part, residual to Repeatability.
% P/T = 100 × k × σ / (USL − LSL).
Destructive / nested study with a numeric covariate (reference measurement, thickness, hardness, …). Each operator measures unique samples. GLM Type I (ANCOVA): covariate first, then Operator, then Sample nested in Operator. Covariate explains material variation; residual → Repeat, Operator → Reprod, Sample → Part-to-Part.
% P/T = 100 × k × σ / (USL − LSL).