Statistics
with the methods
on the page.
For people who want the answer and the receipts, without leaving the workspace.
Quanti is a statistical analysis environment designed for researchers who write up what they did. Every model output ships with the assumptions, the residuals, and the citation block ready to drop into the methods section, not buried in a separate plot window.
Numbers and the
argument they make.
Stats software outputs a number and stops. Quanti outputs the number, the model assumptions, the diagnostic plots, and the prose, ready to paste into a methods section. The whole audit trail, on one page.
Run the right test.
Pick the question, not the procedure. Quanti looks at your data shape (paired? skewed? heteroscedastic?) and recommends the right test, with the alternatives one click away. Every recommendation cites the rule that surfaced it.
Methods, not screenshots.
Output is structured. Each result has a publish-ready methods paragraph in APA-7, MLA or AGLC4 style, with figure callouts that auto-update as you edit. Drop it into Word, Pages, Quarto or LaTeX without re-typing.
An audit trail.
Every run is versioned with its inputs, model, seed, and library version. Hand a Quanti project to a co-author or replication reviewer and they reconstruct your result down to the floating-point digit.
Three views of one regression.
A multiple regression on the Per Capita longitudinal cost-of-living dataset, seen as a model run, a methods card, and a data inspector.
Estimate Std.Err t Pr(>|t|) (Intercept) 0.5142 0.0218 23.6 < .001 *** income_log -0.0867 0.0044 -19.7 < .001 *** household_size 0.0094 0.0021 4.5 < .001 *** tenure[mortgage] -0.0312 0.0061 -5.1 < .001 *** tenure[private rent] 0.0721 0.0058 12.4 < .001 *** tenure[social rent] 0.0418 0.0083 5.0 < .001 *** region[NSW] 0.0186 0.0049 3.8 < .001 *** region[VIC] 0.0134 0.0051 2.6 * .009 … (5 more) R² = 0.412 Adj. R² = 0.408 F(11, 4220) = 268.4 p < .001 RMSE = 0.063 AIC = -11244 BIC = -11168 n = 4,232 ✓ assumptions met
OLS appropriate. linearity ok homoscedastic vif 2.4 income
A multiple linear regression… tab over to write.
Multiple linear regression of housing share on income, household size, tenure and region.
We fitted an ordinary least-squares regression of the share of household expenditure devoted to housing on log-transformed equivalised income, household size, tenure type (owner-outright as reference) and state of residence (ACT as reference) on n = 4,232 households drawn from the Per Capita Cost-of-Living Longitudinal Survey, waves 2019–2024.
The model explained 41.2% of variance in housing share (R² = .412, adj. R² = .408, F(11, 4220) = 268.4, p < .001, RMSE = 0.063). Income (β = −0.087, p < .001) and private-rental tenure (β = 0.072, p < .001) were the largest predictors. Assumption checks found residuals approximately normal (Shapiro–Wilk W = 0.997), homoscedastic (Breusch–Pagan p = .31) and free of multicollinearity above conventional thresholds (max VIF = 2.4).
, auto-generated by Quanti v2.8 · methods rule set APA-7 · 2026-03-14T11:44+10
quanti replay run a91f3e2
| id · int | year · int | income_log · num | housing_share · num | household_size · int | tenure · cat | region · cat | employment · cat |
|---|---|---|---|---|---|---|---|
| 00214 | 2019 | 10.84 | 0.281 | 2 | owner_outright | NSW | retired |
| 00215 | 2019 | 11.22 | 0.318 | 4 | mortgage | VIC | ft_employed |
| 00216 | 2019 | 10.41 | 0.412 | 1 | private_rent | NSW | pt_employed |
| 00217 | 2019 | 10.96 | 0.346 | 3 | mortgage | QLD | ft_employed |
| 00218 | 2019 | 9.88 | 0.488 | 2 | social_rent | SA | unemployed |
| 00219 | 2019 | 11.04 | 0.302 | 4 | owner_outright | VIC | ft_employed |
| 00220 | 2019 | 10.62 | 0.376 | 2 | private_rent | NSW | ft_employed |
| 00221 | 2019 | NA | 0.354 | 3 | mortgage | WA | ft_employed |
| 00222 | 2019 | 10.31 | 0.421 | 1 | private_rent | NSW | student |
| 00223 | 2019 | 10.94 | 0.288 | 5 | mortgage | VIC | ft_employed |
| 00224 | 2019 | 11.41 | 0.241 | 2 | owner_outright | QLD | retired |
| 00225 | 2019 | 10.18 | 0.394 | 4 | private_rent | NSW | pt_employed |
Tests, models, and the methods they ship with.
Forty-two procedures, every one with a methods card, a diagnostic panel, and a citation block. Validated against R’s reference implementation before each release.
OLS, GLM, mixed effects
Linear, logistic, Poisson, ordinal, multinomial, mixed-effects (lme4) and Bayesian (brms) models. One formula bar, the right backend.
Parametric & non-parametric
t-tests, Mann-Whitney, Wilcoxon, Kruskal-Wallis, χ², Fisher, ANOVA / ANCOVA / MANOVA, repeated-measures, post-hoc with Tukey / Bonferroni / Holm.
Diagnostics, on every run
Q-Q, residuals, Cook’s D, leverage, VIF, Shapiro-Wilk, Breusch-Pagan, Durbin-Watson, surfaced beside the result, not a separate window.
Right-test recommender
State the question. Quanti reads your data shape and proposes the test, with cited reasoning, and the alternatives a click away.
Publish-ready prose
APA-7, MLA, AGLC4, Vancouver. Citations auto-attached; figure callouts auto-numbered. Drop into Word, Pages, Quarto or LaTeX.
A-priori & post-hoc power
G*Power-equivalent calculations: required n, observed power, effect-size confidence intervals. Plotted across a parameter range.
SPSS, Stata, R, CSV
Native readers for .sav, .dta, .rdata, .csv, .xlsx, Parquet, Arrow. Variable labels, value labels and formats round-trip cleanly.
Run hashing & replay
Every run is hashed with inputs, seed and library versions. quanti replay run a91f3e2 reconstructs your result exactly.
Open formats throughout
Project files are plain JSON + CSV. No vendor lock-in. Open the project in any text editor; everything important is human-readable.
From CSV to methods section.
How a dataset moves from import to a publishable methods paragraph, with the receipts attached.
Bring the dataset in.
Drop your .sav, .dta, .csv or Parquet onto the workspace. Variable types and labels are inferred; fix or override in seconds.
State the question.
“Does income predict housing share once tenure is held constant?” The recommender proposes the test, cites the rule, and offers the alternatives.
Fit the model.
One formula bar. Diagnostics surface beside the result. Assumption violations surface as warnings, with the robust alternative one click away.
Auto-draft the methods.
The methods card produces APA-7, MLA, AGLC4 or Vancouver prose. Citations attach automatically; figure callouts auto-number.
Hash, archive, hand off.
Every run versioned with seed and library hashes. Hand a Quanti project to a co-author and they reconstruct your result to the digit.