Word
non identifiability
Definition
n. a mathematical problem where completely different internal settings produce the exact same observable data, making it impossible to know which setting is real.
n. a property of a statistical model or structural equation where multiple distinct parameter configurations induce the exact same probability distribution over observables.
Examples
“If you only know the total cost was ten dollars, non-identifiability prevents you from knowing the prices of the two individual items.”
“In causal inference, non-identifiability arises when unmeasured confounders make it impossible to isolate the true treatment effect from observational data alone.”
“Deep generative latent variable models suffer from structural non-identifiability without auxiliary supervision or inductive assumptions on the latent prior.”
Examples
simple
“If you only know the total cost was ten dollars, non-identifiability prevents you from knowing the prices of the two individual items.”
contextual
“In causal inference, non-identifiability arises when unmeasured confounders make it impossible to isolate the true treatment effect from observational data alone.”
complex
“Deep generative latent variable models suffer from structural non-identifiability without auxiliary supervision or inductive assumptions on the latent prior.”
Etymology
From non- + identifiability, derived from identify (Late Latin identificare, from identitas, 'sameness'). Formalized in econometrics and statistics in the mid-20th century.
Etymology adapted from Wiktionary, available under CC BY-SA 4.0.
Domains
Scan code
englishreference.com/q/non-identifiability