Overview
Survlytics provides explicit two-sided tests and procedure-specific effect sizes rather than one generic comparison routine.
Theoretical foundation
t statistics standardise mean differences by estimated sampling error. ANOVA partitions variation or uses Welch weighting. Pearson chi-square compares observed and expected counts; Fisher exact enumerates 2 × 2 table probabilities conditional on margins.
Key equations
t=\frac{\bar{x}_1-\bar{x}_2}{\sqrt{s_1^2/n_1+s_2^2/n_2}}Algorithm and computation
SciPy supplies t and exact distributions. Student’s test pools variances; Welch uses Satterthwaite degrees of freedom. Paired tests operate on complete differences. Conventional ANOVA reports eta² and omega²; Welch ANOVA is calculated from inverse-variance group weights. Chi-square has no continuity correction; sparse 2 × 2 tables can automatically use two-sided Fisher exact inference.
Required data
Numeric outcomes and valid groups for t/ANOVA; paired numeric measurements for paired t; two categorical variables for chi-square/Fisher.
Assumptions
Independence follows the design. Conventional tests require suitable residual distributions; Student/standard ANOVA also require equal variances. Fisher exact is limited to 2 × 2 tables.
Available settings
Student or Welch; conventional or Welch ANOVA; automatic, chi-square or Fisher categorical method; confidence level.
Worked example
A synthetic two-group comparison reports the mean difference, Welch degrees of freedom, 95% interval, Cohen d and Hedges g.
Limitations and non-claims
No repeated-measures ANOVA, nonparametric group tests, post-hoc pairwise procedure, survey weighting or multiplicity correction is implemented.
Validation notes
Formula-level outputs and edge cases are covered by repository numerical tests.
How Surv.lytics computes Categorical inference
{'method': 'auto'}How Surv.lytics computes Independent t test
{'method': 'welch'}How Surv.lytics computes One-way ANOVA
{'method': 'welch'}How Surv.lytics computes Paired t test
{}Was this article helpful?
Your response helps the methodology team improve this documentation.