For each , choose the continuous probability density functions
The second is at least and integrates to one. Their expected values differ by
Writing gives
The supplied product bound, also obtained from the chi-squared divergence of product measures, implies
For the final elementary estimate, compare the exponential series with the geometric series: for . Thus the overlap of the product laws is at least . Part (b) now gives the explicit mean-estimation minimax lower bound for continuous densities
So one may take , uniformly for all . The alternatives are allowed to depend on , because the minimax risk takes a supremum over all continuous probability density functions separately at each sample size.
Apply the preceding argument to the two product observation laws and the real-valued estimator . Alternatively, directly use
The average of the two mean squared errors is at least times the overlap of the two joint probability density functions. Their overlap equals
Taking the infimum over estimators therefore yields
Here is the minimax risk for this functional. The factor is the total variation distance, so the same metric squared-loss two-point bound applies to a functional even when distinct density parameters have the same functional value.
Fix any estimator and put , and . The triangle inequality for the metric gives . Hence
The larger of the two risk functions dominates their average. Using nonnegativity to replace both probability density functions by their minimum,
The right side is independent of the estimator, so taking the infimum proves
This metric squared-loss two-point bound uses overlap of the observation laws to quantify how hard it is to distinguish the two separated parameters.
For the bounded density tilt , differentiate the polynomial in :
Since every score function is centered, the centered representer is
A bounded baseline makes this a bounded mean-zero function, hence an element of . Thus it is the efficient influence function for the density fourth-power functional relative to the regular bounded density tilts. Its squared L2 norm is .
A bounded baseline alone does not make this functional differentiable along every quadratic-mean differentiable path. The derivative above is the intended regular-path answer. To see the need for the qualification, let and put on , extended by zero outside. For , define
Each is a nonnegative continuous probability density function, because its narrow bump has mass . Moreover,
It is therefore a differentiable-in-quadratic-mean path with zero score function. Yet
The density fourth-power functional is not even continuous along this statistical path, although every is individually bounded. Consequently no efficient influence function represents derivatives over the unrestricted class of all such paths.
One sufficient additional condition is a common bound for all small . Taylor expansion then bounds the fourth-power remainder by , while
Together with the quadratic-mean to L1 density derivative, this gives . Under this local bound, or when the chosen statistical paths are the bounded density tilts, the boxed canonical gradient is fully justified. The counterexample is a spike obstruction to density-power differentiability.
For a pathwise differentiable statistical functional, an influence-function representer is a mean-zero function such that every admissible score function satisfies . The efficient influence function, also called the canonical gradient, is the unique such representer in the statistical tangent space. Equivalently, it is the orthogonal projection of any representer onto that statistical tangent space. The Pythagorean theorem in an inner-product space shows that it has the smallest squared L2 norm among all representers.
Here the statistical tangent space is all of . To verify the closure explicitly, take , truncate it to , and set . Then is bounded and centered, and in , by dominated convergence and the Cauchy-Schwarz inequality. Part (c) supplies a representer already in this space. Hence
Its variance is .
The closure must be taken in the density-weighted space . An unweighted reading of in the printed hint is false. For example, when , the function has and , so bounded centered truncations converge to it in ; nevertheless . This illustrates density of bounded centered scores and fixes the measure in the closure statement.
Consider any differentiable-in-quadratic-mean path with score function . Put , where . The quadratic-mean to L1 density derivative follows from
By the Cauchy-Schwarz inequality,
Since is a bounded function, multiplying this L1 norm bound by proves
The last equality uses . The derivative is a bounded linear functional of , so the required pathwise differentiability of a statistical functional holds, in particular relative to the statistical tangent set from part (b). Its derivative is
For the explicit bounded density tilts in part (b), this derivative is also obtained by direct integration, with no remainder term.
Let be a bounded function with . A bounded density tilt realizes this direction:
These are nonnegative probability density functions, since . The uniform Taylor expansion of the square root gives
The squared L2 norm of this remainder is , since is bounded. Thus the statistical path is differentiable in quadratic mean with score function .
Conversely every score function is centered by the mean-zero score identity under quadratic-mean differentiability. Choosing all these bounded density tilts therefore gives the statistical tangent set
This is a valid choice of statistical tangent set; it does not assert that every possible score function in the unrestricted density model is bounded.
Choose a collection of statistical paths through that are differentiable in quadratic mean. A statistical tangent set at is the set of their score functions. In particular each member belongs to , by the mean-zero score identity under quadratic-mean differentiability.
A statistical tangent set records which first-order directions the chosen statistical paths can realize. Its closed linear span in L2 space is the statistical tangent space. The tangent set consists of attainable scores; the tangent space also includes their linear combinations and limits.
The map is a bounded linear functional on L2 space, since by the Cauchy-Schwarz inequality. Its kernel, the mean-zero L2 space, is therefore closed. Every nuisance score function and the parametric score function are centered, so both and are centered. Hence
Moreover, the efficient score belongs to the orthogonal complement of the nuisance tangent space. Writing gives
This is the efficient-score projection identity; it remains valid when the efficient information is zero.
Fix . Let be the nuisance tangent space: the closed linear span in of score functions of statistical paths that vary only the nuisance parameter . By the preceding argument, is contained in the mean-zero L2 space .
Let denote orthogonal projection onto this closed subspace of a Hilbert space. The efficient score and scalar efficient information are
The efficient score is the component of the parametric score function that cannot be reproduced by changing the nuisance parameter. The efficient information is its squared L2 norm; it can be zero, so positivity must not be assumed in the definition.
Put . By differentiability in quadratic mean, and
in L2 space. The second limit follows because the first gives .
Both probability density functions integrate to one. Consequently the L2 inner product satisfies
The Cauchy-Schwarz inequality permits passage to the limit in this inner product, giving
Thus every score function is a mean-zero function. This proof uses square-root differentiability in quadratic mean, without requiring differentiation under the density integral.
A statistical path through is a family , defined on an open interval containing zero, with . Write and . Its differentiability in quadratic mean at zero means that there is a square-integrable function such that
The function is the score function of the statistical path. Thus it is the derivative of the square-root probability density function, multiplied by two and divided by where . The score function is determined only -almost everywhere. The displayed definition also controls any probability mass entering a region where ; a pointwise derivative of the log probability density function alone would not do that. The required derivative is in the square-root-density norm.
One precise relation is to radial SLE with parameter . Its orientation is from the boundary point to the interior target . With conformal-radius time it is defined by
There is a coupling of the conditioned Brownian path with this simple radial curve such that, for each initial curve segment, the first point at which encounters it is its tip. More precisely, if , then . Reversing the Brownian path turns these first-hit times into the last-visit times that characterize loop-erasure of planar Brownian motion. Thus
This is an existence statement for a coupling, not a pointwise equality of the two processes or a claim that a naive finite-loop deletion algorithm applies to Brownian motion.
Here is the martingale mechanism behind the coupling. Put and
This is the positive slit-domain Poisson kernel at the tip, normalized by . Writing , the radial equation gives
The drift of is , so is a local martingale at .
Start independent stopped and radial SLE2, and weight their joint laws, before meeting, by
For fixed this is an SLE local martingale. For fixed it is a Brownian-transform local martingale, because . Since , localized two-parameter weighting preserves both original marginals. For a fixed slit, the changed Brownian law is the transform by , so it hits that slit at the pole, its tip. Compatible stopped couplings, followed by exhaustion, give the first-hit-at-tip property. This completion is the substantive continuum coupling step. Discrete loop-erased random walks and their radial-SLE2 limit give a parallel approximation picture.
For example, in this coupling avoidance by implies avoidance by , since the radial curve is contained in the Brownian trace. The probability from part (b) therefore supplies a lower bound . The two avoidance laws are not identical: unweighted simple radial does not have the Brownian radial-restriction property. This is a way to transfer path-containment information while retaining the distinct geometry and orientation of the two random curves.
Here the map direction is opposite to Question 2: write , with and . The separation condition makes agree with near . Schwarz reflection across that analytic boundary arc therefore defines the boundary derivative, whose modulus is positive.
For ordinary Brownian motion started at , let be its first exit from . Take sufficiently short arcs around . They are also arcs of . The event that stays in until its disc exit and exits in is exactly the event . Therefore the conditional avoidance probability is
Conformal invariance of planar Brownian motion gives . At the disc center harmonic measure is normalized arc length, so the ratio tends to .
To justify that this limit is the avoidance probability for the point-conditioned diffusion, one may use the stopped transform directly. Its probability of reaching a smooth boundary arc at before any other boundary of is the mass at that pole in the transformed harmonic measure. All other exit points have weights and represent failure. Equivalently integrate the Poisson-kernel density at for and divide by that for . This gives the boundary-density ratio above, without assuming that a full-path avoidance event is a continuity set for arbitrary weak convergence. The kernels transform by the boundary Jacobian:
Consequently
There is no extra factor of . The starting point is fixed and the conditioning concerns boundary harmonic-measure density. In this normalization the boundary derivative is a positive real, so it may also be written . It is at most one by the probability interpretation. Given avoidance, the mapped path is the same conditioned Brownian motion up to conformal time change; this is radial restriction for a conditioned Brownian path.
Let be the unit disc and normalize its positive Poisson kernel at the boundary point by
It is harmonic in . Let be the killed Brownian transition density, for planar Brownian motion with generator . The Doob h-transform has sub-Markov transition density
Its missing mass represents paths already absorbed at the distinguished boundary point. Equivalently, for a stopping time before exit from a compact subdomain, its law is weighted relative to ordinary killed Brownian motion by . These stopped laws are consistent and define a diffusion up to its lifetime. Its generator is
This is a rigorous Brownian motion conditioned to exit at a boundary point, not conditioning on an event of positive probability. Exhaustion of shows that its terminal boundary limit is ; its lifetime is finite. For example its expected lifetime from is , which is finite: the Green function vanishes linearly near the smooth boundary and cancels the Poisson-kernel singularity there, while its logarithmic singularity at is integrable.
One can also condition ordinary Brownian motion on exiting through an arc about , and then let its length decrease to zero. The conditional harmonic functions converge locally uniformly to . Thus their stopped conditional laws converge to this same diffusion. Both constructions specify the conditioning unambiguously.

Pinned article: Introduction to the OurBigBook Project

Welcome to the OurBigBook Project! Our goal is to create the perfect publishing platform for STEM subjects, and get university-level students to write the best free STEM tutorials ever.
Everyone is welcome to create an account and play with the site: ourbigbook.com/go/register. We belive that students themselves can write amazing tutorials, but teachers are welcome too. You can write about anything you want, it doesn't have to be STEM or even educational. Silly test content is very welcome and you won't be penalized in any way. Just keep it legal!
We have two killer features:
  1. topics: topics group articles by different users with the same title, e.g. here is the topic for the "Fundamental Theorem of Calculus" ourbigbook.com/go/topic/fundamental-theorem-of-calculus
    Articles of different users are sorted by upvote within each article page. This feature is a bit like:
    • a Wikipedia where each user can have their own version of each article
    • a Q&A website like Stack Overflow, where multiple people can give their views on a given topic, and the best ones are sorted by upvote. Except you don't need to wait for someone to ask first, and any topic goes, no matter how narrow or broad
    This feature makes it possible for readers to find better explanations of any topic created by other writers. And it allows writers to create an explanation in a place that readers might actually find it.
    Figure 1.
    Screenshot of the "Derivative" topic page
    . View it live at: ourbigbook.com/go/topic/derivative
  2. local editing: you can store all your personal knowledge base content locally in a plaintext markup format that can be edited locally and published either:
    This way you can be sure that even if OurBigBook.com were to go down one day (which we have no plans to do as it is quite cheap to host!), your content will still be perfectly readable as a static site.
    Figure 5. . You can also edit articles on the Web editor without installing anything locally.
    Video 3.
    Edit locally and publish demo
    . Source. This shows editing OurBigBook Markup and publishing it using the Visual Studio Code extension.
  3. https://raw.githubusercontent.com/ourbigbook/ourbigbook-media/master/feature/x/hilbert-space-arrow.png
  4. Infinitely deep tables of contents:
    Figure 6.
    Dynamic article tree with infinitely deep table of contents
    .
    Descendant pages can also show up as toplevel e.g.: ourbigbook.com/cirosantilli/chordate-subclade
All our software is open source and hosted at: github.com/ourbigbook/ourbigbook
Further documentation can be found at: docs.ourbigbook.com
Feel free to reach our to us for any help or suggestions: docs.ourbigbook.com/#contact