SI
Theorize This
Super Powers
Weird Stuff
Looks Good
100

This process breaks text into smaller units that a language model can process as numerical vocabulary items.

What is tokenization?

100

This is the tentative statement that often contains equality language and is tested against evidence.

What is the null hypothesis?

100

This standardized score tells how many standard deviations an observation is above or below the mean.

What is a z-score?

100

The family of test that uses medians instead of means.

What is Nonparametric tests?

100

This sampling method divides the population into important subgroups and randomly samples from each one.

What is stratified random sampling?

200

This type of AI creates new text, images, code, or other content rather than only assigning a label or score.

What is generative AI?

200

This term describes the practical way an abstract idea such as loyalty, satisfaction, or risk tolerance is turned into a measurable variable.

What is operationalization?

200

This test is appropriate when comparing the same participants’ scores before and after an intervention.

What is a paired t-test?

200

This statistic measures agreement between raters on categorical judgments after accounting for agreement expected by chance.

What is Cohen’s kappa?

200

This visual EDA tool displays pairwise linear relationships among numerical variables in a color-coded matrix.

What is a correlation heatmap?

300

This prompting technique divides a larger analytics task into ordered steps, such as inspect, clean, model, and interpret.

What is prompt chaining?

300

These define where, when, and for whom a theory is expected to apply.

What are boundary conditions?

300

This test compares the means of three or more independent groups on a single continuous outcome.

What is one-way ANOVA?

300

This nonparametric test is commonly used as an alternative to one-way ANOVA when assumptions are not met.

What is the Kruskal-Wallis test?

300

This diagnostic warns that predictors in a regression may be too highly related to one another.

What is Variance Inflation Factor, or VIF?

400

This safety problem occurs when hidden instructions inside retrieved or uploaded content try to override the developer or user’s intended directions.

What is prompt injection?

400

This principle favors the simpler theory when two explanations account for the evidence equally well.

What is parsimony, or Ockham’s Razor?

400

This probability is the chance of correctly detecting a real effect when it exists.

What is statistical power?

400

This nonparametric method evaluates repeated or blocked comparisons across several related conditions.

What is the Friedman test?

400

This sampling method is easy to implement but especially vulnerable to selection bias because participants are chosen by ease of access.

What is convenience sampling?

500

Connects LLM risks to document-based AI workflows.

What is Retrieval-Augmented Generation, or RAG?

500

This type of error occurs when a researcher concludes there is an effect even though the null hypothesis is actually true.

What is a Type I error?

500

Running this model tests 5 treatment groups of a single independent variable against 3 response or dependent variables. 

What is a MANOVA?

500

This diagnostic test for randomness.

What is the Runs test?

500

Use this test when the dependent variable is limited to binary outcomes.

What logistic regression?