Math 496 Homework 2
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
496 Homework 2
Due Tuesday, Feb. 14
Guidelines to Turn the HW in:
(b) The solution to each problem must contain a summary of the main conclusions or final answers, handwritten or typed using either word or R Markdown (or a combination of these).
(b) If you use R to obtain the solutions, explicitly refer to the R output. You also need to include the R code as a separate image or included in the word or Markdown file.
Problems.
1. Let X, X1 , . . . , Xn f and define
γ := 匝 ╱尸 f (x)fˆn (x)dx、= 尸 f (x)匝 ╱fˆn (x)、dx.
Since 匝 ╱fˆn (x)、= 匝(K(x 一 X)) =: fsm (x), we have γ = f (x)fsm (x)dx. In
class, we claimed that
匝 ╱ i1 fˆ(↘i)(Xi )\ = γ,
where fˆ(↘i)(x) is the leave-one-out kernel density estimator. Show the validity of this claim. Hint: First, compute 匝 ╱fˆ(↘i)(Xi )|Xi、.
2. Get the data on fragments of glass collected in forensic work from canvas.
(a) Estimate the density of the first variable (refractive index) using a his- togram and a kernel density estimator. Experiment with 2 or 3 different binwidths and bandwidths.
(b) Now, use cross-validation to choose the amount of smoothing. Add the
confidence intervals fˆn (x) 士 z在 (x) and generate a graph sim-
ilar to the first graph in Section 2.5 of Garc´ıa-Portgu´es’ notes.
3. Exercise 2.13 in “A Short Course on Nonparametric Curve Estimation” by Garc´ıa-Portugu´es.
Remark: Only consider the bimodal density (see Section 2.4 in the notes), and the bandwidth selectors RT (Rule of Thumb), UCV (least-squares cross- validation method as seen in class), and DPI (the Sheather-Jones method).
4. In class, we computed the coverage probabilities of the CIs fˆn (x)士z在 (x) when used to estimate fsm (x) := 匝K(x 一 X). This was done when the band-
width h is set to a fixed value. Carry out the same analysis when the bandwidth h is chosen according to (a) the rule of thumb method, and according to (b) cross-validation. Generate the same graph as in class for these two cases (see graph above Exercise 2.3 in Section 2.5 of Garc´ıa-Portugu´es notes).
5. In class, we computed the coverage probabilities of the CI
fˆn (x) 士 zα/2 ←fˆn (x) K2 (u)du
when applied to estimate both fsm (x) = 匝(Kh (x 一 X) and f (x). We also proposed the following “unbiased” estimator:
f˜n (x; h, ζ) = fˆn (x; h) 一 (fˆn (x; ζh) 一 fˆn (x; h))2
for some fixed ζ > 1.
(a) Compute the coverage probabilities of the following estimator
f˜n (x; h, ζ) 士 zα/2 ←fˆn (x) K2 (u)du
when applied to estimate f (x), and generate a graph similar to that in Section 2.5 of Garc´ıa-Portugu´es (see graph above Exercise 2.3 therein). Choose h according to the RT method and experiment with a few values of ζ and comment on the results.
(b) Repeat (a) with the CI
f˜n (x; h, ζ) 士 zα/2 ←f˜n (x; h, ζ) K2 (u)du
2023-02-11