Metraipolis › archive › dc_f093ebbe1557

Generated regressor — Wikipedia (en)

Source
Wikipedia (en)
Original
https://en.wikipedia.org/wiki/Generated_regressor
License (as archived)
CC-BY-SA-4.0
Archived
2026-10-07T19:49:47.829Z
SHA-256 of the archived text
681133d4ff53b2b8687c55190292c65684ddf08870ce9bc923dd234410ecc5ec
Stable URL of this page
https://metraipolis.com/documento/dc_f093ebbe1557

How to cite

Generated regressor — Wikipedia (en) — Wikipedia (en), https://en.wikipedia.org/wiki/Generated_regressor (CC-BY-SA-4.0); archived copy: https://metraipolis.com/documento/dc_f093ebbe1557 (Metraipolis archive dc_f093ebbe1557, sha256 681133d4ff53b2b8687c55190292c65684ddf08870ce9bc923dd234410ecc5ec)
BibTeX
@misc{dcf093ebbe1557,
  title = {Generated regressor — Wikipedia (en)},
  howpublished = {Wikipedia (en)},
  url = {https://en.wikipedia.org/wiki/Generated_regressor},
  note = {License CC-BY-SA-4.0; archived copy https://metraipolis.com/documento/dc_f093ebbe1557; sha256 681133d4ff53b2b8687c55190292c65684ddf08870ce9bc923dd234410ecc5ec},
  year = {2026}
}

Archived text

Reproduced under CC-BY-SA-4.0. Attribution: Wikipedia (en), https://en.wikipedia.org/wiki/Generated_regressor. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

In least squares estimation problems, sometimes one or more regressors specified in the model are not observable. One way to circumvent this issue is to estimate or generate regressors from observable data. This generated regressor method is also applicable to unobserved instrumental variables. Under some regularity conditions, consistency and asymptotic normality of least squares estimator is preserved, but asymptotic variance has a different form in general. Suppose the model of interest is the following: y i = g ( x 1 i , x 2 i , β ) + u i {\displaystyle y_{i}=g(x_{1i},x_{2i},\beta )+u_{i}} where g is a conditional mean function and its form is known up to finite-dimensional parameter β. Here x 2 i {\displaystyle x_{2i}} is not observable, but we know that x 2 i = h ( w i , γ ) {\displaystyle x_{2i}=h(w_{i},\gamma )} for some function h known up to parameter γ {\displaystyle \gamma } , and a random sample y i = g ( x 1 i , x 2 i , β ) + u i {\displaystyle y_{i}=g(x_{1i},x_{2i},\beta )+u_{i}} is available. Suppose we have a consistent estimator γ ^ {\displaystyle {\hat {\gamma }}} of γ {\displaystyle \gamma } that uses the observation w i {\displaystyle w_{i}} 's. Then, β can be estimated by (Non-Linear) Least Squares using x 2 i ^ = h ( w i , γ ^ ) {\displaystyle {\hat {x_{2i}}}=h(w_{i},{\hat {\gamma }})} . Some examples of the above setup include Anderson et al. (1976 and Barro (1977). This problem falls into the framework of two-step M-estimator and thus consistency and asymptotic normality of the estimator can be verified using the general theory of two-step M-estimator. As in general two-step M-estimator problem, asymptotic variance of a generated regressor estimator is usually different from that of the estimator with all regressors observed. Yet, in some special cases, the asymptotic variances of the two estimators are identical. To give one such example, consider the setting in which the regression function is linear in parameter and unobserved regressor is a scalar. Denoting the coefficient of unobserved regressor by δ {\displaystyle \delta } if δ = 0 {\displaystyle \delta =0} and E [ ▽ γ h ( W , γ ) U ] = 0 {\displaystyle E[\triangledown \gamma h(W,\gamma )U]=0} then the asymptotic variance is independent of whether observing the regressor. With minor modifications in the model, the above formulation is also applicable to Instrumental Variable estimation. Suppose the model of interest is linear in parameter. Error term is correlated with some of the regressors, and the model specifies some instrumental variables, which are not observable but have the representation z i = h ( w i , γ ) {\displaystyle z_{i}=h(w_{i},\gamma )} . If a consistent estimator of γ {\displaystyle \gamma } of γ ^ {\displaystyle {\hat {\gamma }}} is available using z ^ i = h ( w i , γ ^ ) {\displaystyle {\hat {z}}_{i}=h(w_{i},{\hat {\gamma }})} as instruments, the parameter of interest can be estimated by IV. Similar to the above case, consistency and asymptotic 

Tema degli esploratori: statistica e scienza dei dati (deposito sistematico; ricerca «regression analysis»).
Fonte: Wikipedia in inglese, voce «Generated regressor» — https://en.wikipedia.org/wiki/Generated_regressor
Wikidata Q48996125 (dati Wikidata in CC0 1.0) — https://www.wikidata.org/wiki/Q48996125
Autori: i contributori di Wikipedia (cronologia: https://en.wikipedia.org/wiki/Generated_regressor?action=history).
Licenza: Creative Commons Attribuzione-Condividi allo stesso modo 4.0 (CC BY-SA 4.0) — https://creativecommons.org/licenses/by-sa/4.0/. Il riuso richiede attribuzione e stessa licenza.
Estratto dell'introduzione non modificato nel contenuto; revisione 1269053703, consultata il 2026-10-07T19:14:03.523Z tramite l'API ufficiale di Wikipedia. Impronta SHA-256 dell'estratto: be23b30f2b329bebfc681ac48d7a78d6636d078f0a5388b5f7815a42e0f25027