Page 388 - Contributed Paper Session (CPS) - Volume 2
P. 388
CPS1891 Tzee-Ming Huang
In summary, for a smooth , a simple knot selection algorithm using equally
̂
spaced knots like Method 3 is sufficient for giving a satisfactory estimator .
For with a sharp change, a
Algorithm 1 Method 2 Method 3
1 2.76 × 10 1.75 × 10 8.45 × 10
−3
−4
−5
2 0.00149 0.00113 0.000289
3 2.51 × 10 7.69 × 10 8.53 × 10
−5
−4
−4
4 1.70 × 10 1.86 × 10 2.44 × 10
−4
−3
−4
Table 3: Standard deviations of the mean squared errors
more flexible knot selection algorithm such as Algorithm 1 or Method 2, is
needed. Algorithm 1 is more stable than Method 2.
4. Discussion and conclusion
In this article, a knot selection algorithm (Algorithm 1) based on a statistical
test is proposed. Based on the simulation results in Section 3, overall, the
performance of Algorithm 1 is satisfactory. When has a sharp change,
Algorithm 1 outperforms Method 3, which uses equally spaced knots only, and
Algorithm 1 sometimes even performs better than Method 2, which allows for
more flexible knot selection. This satisfactory result is probably due to the fact
that Algorithm 1 is flexible enough to use unequally spaced knots yet not too
flexible to allow for arbitrary knot locations, which makes it more stable than
a knot estimation method such as Method 2.
For future work, a possible direction is to extend Algorithm 1 to the case
where is multi-variate and tensor basis functions are used. Given that
Algorithm 1 does not require searching for best knot locations, it is expected
that the extension to the multivariate case is feasible.
References
1. Kaishev, V. K., Dimitrova, D. S., Haberman, S., and Verrall, R. J. (2016).
Geometrically designed, variable knot regression splines. Computational
Statistics, 31(3):1079{1105.
2. Lindstrom, M. J. (1999). Penalized estimation of free-knot splines. Journal
of Computational and Graphical Statistics, 8(2):333{352.
3. Stone, C. J. (1982). Optimal global rates of convergence for
nonparametric regression. Ann. Statist., 10(4):1040{1053.
4. Stone, C. J. (1994). The use of polynomial splines and their tensor
products in multivariate function estimation. Ann. Statist., 22(1):118{171.
5. Yuan, Y., Chen, N., and Zhou, S. (2013). Adaptive B-spline knot selection
using multi-resolution basis set. IIE Transactions, 45(12):1263{1277.
6. Zhou, S. and Shen, X. (2001). Spatially adaptive regression splines and
accurate knot selection schemes. Journal of the American Statistical
Association, 96:247{259.
377 | I S I W S C 2 0 1 9