Data Types Analysis Choice

Sam

Where it fits conceptually
This table belongs in the foundational part of any statistical methods overview, right after you introduce types of variables (nominal vs. interval/continuous).

For the table below: Verify that underlying assumptions (normality, homoscedasticity, independence) are met, or consider non-parametric or alternative approaches if they are not.

DV	IV	Typical Methods / Models	Notes / Use Cases
Nominal (2 categories)	Nominal (2 categories)	- 2×2 contingency table (Chi-square test) - Fisher’s exact test	For two categorical variables (often “success/failure”). Fisher’s exact test is used for small sample sizes.
Nominal (>2 categories)	Nominal (>2 categories)	- R×C contingency table (Chi-square test) - Fisher’s exact test (small samples)	For two categorical variables with multiple levels (e.g., a 3×4 table for “brand preference” vs. “region”).
Nominal (2 categories)	Interval/Continuous or Mixed (could include nominal)	- Logistic regression - Probit regression - Discriminant analysis	Binary DV (yes/no) predicted by numeric or categorical IVs.
Nominal (>2 categories)	Interval/Continuous or Mixed	- Multinomial logistic regression - Ordinal logistic regression (if categories have order) - Classification trees	Categorical DV with more than two classes (e.g., “Type A/B/C” outcome) predicted by numeric/categorical variables.
Ordinal (ordered categories)	Interval/Continuous or Mixed	- Ordinal logistic regression - Nonparametric tests (Kruskal–Wallis for group comparisons)	Rank-based or ordered responses (e.g., Likert scale 1–5) predicted by numeric or categorical IVs.
Interval/Continuous (numeric)	Nominal (2 categories)	- Two-sample t-test (independent) - Paired t-test (if repeated measures) - Wilcoxon rank-sum (nonparametric)	Compare means of a continuous DV across two groups (e.g., “test vs. control”), or the same group measured twice (paired).
Interval/Continuous (numeric)	Nominal (>2 categories)	- One-way ANOVA - Factorial ANOVA (≥2 factors) - Repeated-measures ANOVA - Kruskal–Wallis (nonparametric)	Compare means of a continuous DV across multiple groups/factors (e.g., “3 treatment groups” or “2 factors × 2 levels”).
Interval/Continuous (numeric)	Interval/Continuous (numeric)	- Correlation (Pearson, Spearman) - Simple linear regression - Multiple linear regression - ANCOVA (mix)	Relate continuous DV to one or more numeric IVs. Correlation measures association; regression models the numeric relationship.
Count (non-negative integers)	Interval/Continuous or Mixed	- Poisson regression - Negative binomial regression - Chi-square goodness-of-fit (for distributions)	DV is a count (e.g., “number of defects”). Poisson or negative binomial are typical if over-dispersion is an issue.
Time-to-event (a.k.a. “survival”)	Interval/Continuous or Mixed	- Kaplan–Meier survival curves (for univariate factors) - Cox proportional hazards regression - Parametric survival models	DV measures time until an event (e.g., “time to failure”)—common in biomedical, reliability, and event-history analyses.