As we know that we can perform a Bias Variance decomposition of an Estimator with MSE as loss function and it will look like below:

$$\operatorname{MSE}(\hat{\theta}) = \operatorname{tr}(\operatorname{Var}[\hat{\theta}]) + (\|{\operatorname{Bias}[\hat{\theta}]}\|)^2$$

Similarly, if we want to perform a Bias Variance decomposition of an predictor with MSE as a loss function then we it will look like:

$$\operatorname{MSE}(\hat{y}\mid X) = \operatorname{Var}[\hat{y}] + (\|{\operatorname{Bias}[\hat{y}]}\|)^2 + \sigma_{\varepsilon}^2 $$

I am more curious to to know the philosophy to break down a estimator or a predictor into Variance and Bias term. Why not some other terms? It is more of a broad question of why we can think of breaking estimators and predictors into this form.

Just thinking aloud we can break a predictor into known distribution plus and error term or an estimator into an know distribution of the sample and an error term.

Please do correct me if I have some misunderstanding in terms of my thought process.

The paper which triggered this question in my head (Bit unrelated): https://faculty.wharton.upenn.edu/wp-content/uploads/2012/04/Strong.pdf


  • Edit 1: Predictor Error with $\sigma_\varepsilon^2$
  • Edit 2: Updated reference paper
  • $\begingroup$ Such a decomposition is a logical and nice consequence of the MSE. Indeed bias an variance of an estimator are atomic or elementary properties that we wish our estimator satisfies. $\endgroup$
    – utobi
    Commented Jan 24, 2023 at 20:13
  • $\begingroup$ "I am more curious to to know the philosophy to break down a estimator or a predictor into Variance and Bias term." I've reread this several times, and I can't understand it. $\endgroup$
    – AdamO
    Commented Jan 24, 2023 at 20:20
  • $\begingroup$ As utobi said there are the elementary properties... @AdamO - If we want to make an estimator/predictor we need the atomic or elementary properties which will make the final estimaor... I am willing to know why we are sticking to bias and variance and what is the reasoning or thought process behind it and not something else. $\endgroup$
    – Rehan Guha
    Commented Jan 24, 2023 at 20:27
  • 1
    $\begingroup$ $\hat{y}$ is a function of parameter estimates, in my experience it's kind of a trivial exercise to explore the properties of fitted values as a function of the properties of the regressors. But model specification, omitted variable bias, etc. are what you'd need to look into and it's a whole different topic/question. $\endgroup$
    – AdamO
    Commented Jan 24, 2023 at 23:35
  • 1
    $\begingroup$ Neither does the answer by @AdamO other than describing it as an elegant result. I think the question "why not do something another way" without proposing what shape this other way might take has no good answer. $\endgroup$
    – dipetkov
    Commented Jan 25, 2023 at 9:53

1 Answer 1


Bias and variance are elementary properties of estimators, and they're usually introduced to early statistics students because they're well understood conceptually, and one can study the properties of the quite restricted class of unbiased estimators - namely the Cramer-Rao bound, sufficiency, asymptotic relative efficiency, etc.

The fact that squared bias and variance result as a decomposition of squared error is, if anything, an elegant result. The number of questions that immediately follow are too numerous to count, is squared error loss the right loss? Under what conditions is it optimal? Does a similar result exists for other loss functions? To paraphrase Paul Erdos, "Anyone can think of an interesting problem."

The first year theory approach has its problems too. Consider Hodge's superefficient estimator. It is unbiased and it beats the Cramer-Rao bound. But it turns out that the estimator is not regular.

More broadly, once we start considering biased estimators, we have a much broader class of estimators with different optimality properties to consider. Concepts like admissibility, minimax, penalized or bounded loss, etc. give rise to other popular estimators as solutions to particular problems, particularly Bayes estimators, ridge estimators, and so on. These concepts would be covered in a second year statistics or probability theory class, from such texts as Ferguson "A Course in Large Sample Theory", Lehmann Casella "Theory of Point Estimation", or Wassermans' "All of Nonparametric Statistics".


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.