Metamodels are an important tool in simulation analysis as they can provide insight about the behavior of the simulation response. Modeling the response with low-degree polynomial segments allows the identification of different behavior zones and the parameters still have relation with the physical world. The purpose of this paper is to extend the use of segmented polynomial functions for simulation metamodeling, where the segments have at most identical value and slope at the breaks. Our approach is to build segmented polynomials metamodels where the hypothesis of degree and continuity of splines are less exigent, allowing more flexibility of the approximation. When breaks are known, constrained least squares are used for metamodel estimation, taking into account the linear formulation of the problem. If breaks have to be estimated, the unconstrained nonlinear regression theory is used, when it can be applied. Otherwise, the estimation is performed using an iterative algorithm which is applied repeatedly in a cyclic manner for estimating the breaks, and jackknifing yields the confidence intervals.
GallantARFullerWA. Fitting segmented polynomial regression models whose join points have to be estimated. Eur J Oper Res1973; 68: 144–147.
19.
SeberGAFWildCJ. Nonlinear regression. New York: John Wiley and Sons, 1989.
20.
BaconDWattsD. Estimating the transition between two intersecting straight lines. Biometrika1971; 58: 525–534.
21.
GriffithsDMillerA. Hyperbolic regression – a model based on two-phase piecewise linear regression with a smooth transition between regimes. Comm Stat B Simulat Comput1973; 2: 561–569.
22.
DaughetyAFTurnquistMA. Simulation optimization using response surfaces based on spline approximations. In: HighlandH, editor. Proceedings of the Winter Simulation Conference. 1978, IEEE, Piscataway, NJ. pp. 183–193.
23.
KeysACReesLP. A sequential-design metamodeling strategy for simulation optimization. Comput Oper Res2004; 31: 1911–1932.
24.
WangHPasupathyRSchmeiserB. Integer-ordered simulation optimization using R-SPLINE: Retrospective search with piecewise-linear interpolation and neighborhood enumeration. ACM Trans Model Comput Simul2013; 23: 1–17.
25.
KangBGSeoKMKimTG. Machine learning-based discrete event dynamic surrogate model of communication systems for simulating the command, control, and communication system of systems. Simulation2018; 95: 673–691.
KleijnenJPCSanchezSMLucasTW, et al. A User’s guide to the brave new world of designing simulation experiments. INFORMS J Comput2005; 17: 263–289.
28.
BoydSVandenbergheL. Introduction to applied linear algebra - vectors, matrices, and least squares. Cambridge: Cambridge University Press, 2018.
29.
AmemiyaT. Advanced econometrics. Cambridge, MA: Harvard University Press, 1985.
30.
GallantAR. The theory of nonlinear regression as it relates to segmented polynomial regressions with estimated join points. Institute of Statistics Mimeograph Series, No 925, North Carolina State University, 1974.
31.
FullerWA. Grafted polynomials as approximating functions. Aust J Agr Econ1969; 13: 35–46.
32.
WangJ. Asymptotics of least-squares estimators for constrained nonlinear regression. Ann Stat1996; 24: 1316–1326.
33.
BrentR. Algorithms for minimization without derivatives. Englewood Cliffs, NJ: Prentice-Hall, 1973.
34.
QuenouilleM. Problems in plane sampling. Ann Math Stat1949; 20: 355–375.
35.
TukeyJ. Bias and confidence in not-quite large sample (abstract). Ann Math Stat1958; 29: 614.
36.
KleijnenJPCKarremansPOortwijnW, et al. Jackknifing estimated weighted least squares. Comm Stat Theor Meth1987; 16: 747–764.
37.
DingeçKDAlexopoulosCGoldsmanD, et al. Jackknifed variance estimators for simulation output analysis. In: YilmaziLChanWMoonIRoederTMacalCRossettiM editors. Proceedings of the Winter Simulation Conference. 2015, IEEE, Piscataway, NJ. pp. 459–471.
38.
MillerRG. The Jackknife – A review. Biometrika1974; 61: 1–15.
39.
KleijnenJPCSargentRG. A methodology for fitting and validating metamodels in simulation. Eur J Oper Res2000; 120: 14–29.
40.
KleijnenJPCvan BeersWCM. Application-driven sequential designs for simulation experiments: Kriging metamodeling. J Oper Res Soc2004; 55: 876–883.
41.
SantosMISantosPM. Sequential experimental designs for nonlinear regression simulation metamodels. Simulat Model Prac Theor2008; 16: 1365–1378.
42.
NielsenHBLophavenSNSøndergaardJ. DACE - A Matlab Kriging Toolbox. Richard Petersens Plads, Building 321, DK-2800 Kgs. Lyngby, 2002.
43.
BrantleyMWLeeLHChenCH, et al. An efficient simulation budget allocation method incorporating regression for partitioned domains. Automatica2014; 50: 1391–1400.
