Here, we briefly review a subtlety associated with machine-learning model selection: the fact that the optimal hyperparameters for a model can vary with training set size, $N.$ To illustrate this point, we derive expressions for the optimal strength for both $L_1$ and $L_2$ regularization in single-variable models. We find that the optimal $L_2$ approaches a finite constant as $N$ increases, but that the optimal $L_1$ decays exponentially fast with $N.$ Sensitive dependence on $N$ such as this should be carefully extrapolated out when optimizing mission-critical models.

# Statistics

## Average queue wait times with random arrivals

Queries ping a certain computer server at random times, on average $\lambda$ arriving per second. The server can respond to one per second and those that can’t be serviced immediately are queued up. What is the average wait time per query? Clearly if $\lambda \ll 1$, the average wait time is zero. But if $\lambda > 1$, the queue grows indefinitely and the answer is infinity! Here, we give a simple derivation of the general result — (9) below.

## Improved Bonferroni correction factors for multiple pairwise comparisons

A common task in applied statistics is the pairwise comparison of the responses of $N$ treatment groups in some statistical test — the goal being to decide which pairs exhibit differences that are statistically significant. Now, because there is one comparison being made for each pairing, a naive application of the Bonferroni correction analysis suggests that one should set the individual pairwise test sizes to $\alpha_i \to \alpha_f/{N \choose 2}$ in order to obtain a desired family-wise type 1 error rate of $\alpha_f$. Indeed, this solution is suggested by many texts. However, implicit in the Bonferroni analysis is the assumption that the comparisons being made are each mutually independent. This is not the case here, and we show that as a consequence the naive approach often returns type 1 error rates far from those desired. We provide adjusted formulas that allow for error-free Bonferroni-like corrections to be made.

[edit (7/4/2016): After posting this article, I’ve since found that the method we suggest here is related to / is a generalization of Tukey’s range test — see here.]

## Maximum-likelihood asymptotics

In this post, we review two facts about maximum-likelihood estimators: 1) They are consistent, meaning that they converge to the correct values given a large number of samples, $N$, and 2) They satisfy the Cramer-Rao lower bound for unbiased parameter estimates in this same limit — that is, they have the lowest possible variance of any unbiased estimator, in the $N\gg 1$ limit.

## A review of parameter regularization and Bayesian regression

Here, we review parameter regularization, which is a method for improving regression models through the penalization of non-zero parameter estimates. Why is this effective? Biasing parameters towards zero will (of course!) unfavorably bias a model, but it will also reduce its variance. At times the latter effect can win out, resulting in a net reduction in generalization error. We also review Bayesian regressions — in effect, these generalize the regularization approach, biasing model parameters to any specified prior estimates, not necessarily zero.

This is the second of a series of posts expounding on topics discussed in the text, “An Introduction to Statistical Learning”. Here, we cover material from its Chapters 2 and 6. See prior post here.

## Stochastic geometric series

Let $a_1, a_2, \ldots$ be an infinite set of non-negative samples taken from a distribution $P_0(a)$, and write

$$\tag{1} \label{problem}

S = 1 + a_1 + a_1 a_2 + a_1 a_2 a_3 + \ldots.

$$

Notice that if the $a_i$ were all the same, $S$ would be a regular geometric series, with value $S = \frac{1}{1-a}$. How will the introduction of $a_i$ randomness change this sum? Will $S$ necessarily converge? How is $S$ distributed? In this post, we discuss some simple techniques to answer these questions.

Note: This post covers work done in collaboration with my aged p, S. Landy.

## How not to sort by average rating, revisited

What is the best method for ranking items that have positive and negative reviews? Some sites, including reddit, have adopted an algorithm suggested by Evan Miller to generate their item rankings. However, this algorithm can sometimes be unfairly pessimistic about new, good items. This is especially true of items whose first few votes are negative — an issue that can be “gamed” by adversaries. In this post, we consider three alternative ranking methods that can enable high-quality items to more-easily bubble-up. The last is the simplest, but continues to give good results: One simply seeds each item’s vote count with a suitable fixed number of hidden “starter” votes.

(more…)

## Multivariate Cramer-Rao inequality

The Cramer-Rao inequality addresses the question of how accurately one can estimate a set of parameters $\vec{\theta} = \{\theta_1, \theta_2, \ldots, \theta_m \}$ characterizing a probability distribution $P(x) \equiv P(x; \vec{\theta})$, given only some samples $\{x_1, \ldots, x_n\}$ taken from $P$. Specifically, the inequality provides a rigorous lower bound on the covariance matrix of any unbiased set of estimators to these $\{\theta_i\}$ values. In this post, we review the general, multivariate form of the inequality, including its significance and proof.

(more…)

## Mathematics of measles

Here, we introduce — and outline a solution to — a generalized SIR model for infectious disease. This is referenced in our following post on measles and vaccination rates. Our generalized SIR model differs from the original SIR model of Kermack and McKendrick in that we allow for two susceptible sub-populations, one vaccinated against disease and one not. We conclude by presenting some python code that integrates the equations numerically. An example solution obtained using this code is given below.

(more…)