Local Moment for Heterogenous Regression and its Application to Estimation and Local Change-Point Detection

Abstract

In this paper, we define a local moment for heterogenous regression and employ it to construct simplified methods for parameter estimation and local change-point detection. The new idea is motivated by our findings that the moment conditions of the model contain the information of homogenous parameter and the subgroup-averages of the heterogenous parameters. Thus we directly use the moment conditions to construct the estimator of the homogenous parameter, and identify the subgroup-averages of the heterogenous parameters. The resulting estimator for homogeneous parameter has a simple expression, and is adaptive to various sizes of subgroups of heterogenous parameters. Based on the subgroup moment estimators, the change-point detections can be achieved by local diagnostic methodology and local informational strategy. The methods are much easier than the existing methods, and are adaptive to various conditions. Our approaches are further illustrated via simulation studies and are applied to non-performing loan model.

Keywords

Heterogeneity / Moment condition / Change-point detection / Adaptability / Consistency / 62J05 / 62F03 / 62H30

References

[1] Arlot, S., Celosse, A., Harchaoui, Z.: Kernel change-point detection. Available at (2012) arXiv:1202.3878

[2] BaiJ, PerronP. Estimating and testing linear models with multiple structural changes. Econometrica, 1998, 66: 47-78

[3] BaiJ, PerronP. Computation and analysis of multiple structural change models. J. Appl. Econom., 2003, 18: 10-22

[4] BraunJV, BraunRK, MällerHG. Multiple changepoint fitting via quasi likelihood, with application to DNA sequence segmentation. Biometrika, 2000, 87: 301-314

[5] Chen, J., Gupta, A.K.: Parametric Statistical Change Point Analysis with Applications to Genetics, 2nd Edn, p. 2012. LLC, Medicine and Finance. Springer Science+Business Media (2000)

[6] ChenJ, GuptaAK. Testing and locating variance change points with application to stock prices. J. Amer. Statist. Assoc., 1997, 92: 739-747

[7] ChenH, ZhangNR. Graph-based change-point detection. Ann. Stat., 2012, 43: 139-176

[8] ChiE, LangeK. Splitting methods for convex clustering. J. Comput. Graph. Stat., 2016, 24: 994-1013

[9] ChipmanH, GeogeE, McCullochR. Bayesian treed models. Mach. Learn., 2002, 48: 303-324

[10] Dette, H., Pan, G., Yang, Q.: Estimating a change point in a sequence of very high-dimensional covariance matrices. J. Amer. Statist. Assoc. (to appear) (2020)

[11] DetteH, GösmannJ. A likelihood ratio approach to sequential change point detection for a general class of parameters. J. Amer. Statist. Assoc., 2020, 115: 1361-1377

[12] FeterP. On asymptotic distribution theory in segmented regression problems-identified case. Ann. Stat., 1975, 3: 49-83

[13] FeterP. The log likelihood ratio in segmented regression. Ann. Stat., 1975, 3: 84-97

[14] GordonRD. Values of mills’ratio of area to bounding ordinate and of the nor- mal probability integral for large values of the argument. Ann. Math. Stat., 1941, 12: 364-366

[15] GramacyR, LianH. Gaussian process single-index models as emulators for computer experiments. Technometrics, 2012, 54: 30-41

[16] HaoN, NiuYS, ZhangHP. Multiple change-point detection via a screening and ranking algorithm. Stat. Sin., 2013, 23: 1553-1572

[17] HarchaouiZ, Lévy-LeducC. Multiple change-point estimation with a total variation penalty. J. Amer. Statist. Assoc., 2010, 105: 1480-1493

[18] HawkinsDM. Fitting multiple change-point models to data. Comput. Statist. Data Anal., 2001, 37: 323-341

[19] KillickR, FearnheadP, EckleyIA. Optimal detection of changepoints with a linear computational cost. J. Amer. Statist. Assoc., 2012, 107: 1590-1598

[20] LavielleM. Using penalized contrasts for the change-points problems. Signal Process, 2005, 85: 1501-1510

[21] LermenPM. Fitting segmented regression models by grid search. Appl. Statist., 1980, 20: 77-84

[22] LiKC, LueHH, ChenCH. Interactive tree-structured regression via principal Hessian directions. J. Am. Stat. Assoc., 2000, 95: 547-560

[23] LinL, ShiYF, WangX, YangSZ. $k$-sample upper expectation linear regression. J. Statis. Plan. Inference, 2016, 170: 15-26

[24] LinL, DongP, SongY, ZhuL. Upper expectation parametric regression. Stat. Sin., 2017, 27: 1265-1280

[25] LinL, LiuYX, LinC. Mini-max-risk and mini-mean-risk inferences for a partially piecewise regression. Statistics, 2017, 51: 745-765

[26] LiuB, ZhouC, ZhangX, LiuY. A unified data-adaptive framework for high dimensional change point detection. J. R. Statist. Soc. B, 2020, 82: 933-963

[27] Ma, S., Huang, J.: Estimating subgroup-specific treatment effects via concave fusion. (Manuscript) (2016)

[28] MaS, HuangJ. A concave pairwise fusion approach to subgroup analysis. J. Am. Statist. Assoc., 2017, 112: 410-423

[29] McgeeVE, CarletonWT. Piecewise regression. J. Am. Statist. Assoc., 1970, 65: 109-1124

[30] NiuYS, ZhangHP. The screening and ranking algorithm to detect DNA copy number variations. Ann. Appl. Statist., 2012, 6: 1306-1326

[31] On segmented multivariate regression. Liu, J., Shiying, Wu., S., Zidek, J. W. Statistica Sinica, 1997, 7: 497-525

[32] OsipovLV, PetrovVV. Short communications on an estimate of the remainder term in the central limit theorem. Theory Prob. Appl., 1967, 12: 281-286

[33] QuandtR. The estimation of the parameters of a linear regression system obeying tow separate regimes. J. Am. Statist. Assoc., 1958, 53: 873-880

[34] Rigaill, G.: Pruned dynamic programming for optimal multiple change-point detection. Available at (2010) arXiv:1004.0887

[35] SrivastavaMS, WorsleyKJ. Likelihood ratio tests for a change in the multivariate normal mean. J. Am. Statist. Assoc., 1986, 81: 199-204

[36] TibshiraniR. Regression shrinkage and selection via the lasso. J. R. Stat. Soc. Ser. B, 1996, 58: 267-288

[37] TishlerA, ZangI. A maximum likelihood method for piecewise regression models with continuous dependent variable. Appl. Statist., 1981, 30: 116-124

[38] TishlerA, ZangI. A new maximum likelihood algorithm for piecewise regression. J. Am. Statist. Assoc., 1981, 76: 980-987

[39] WangT, XiaY. A piecewise single-index model for dimension reduction. Technometrics, 2014, 56: 312-324

[40] WangT, ZhangJ, LiangH, ZhuL. Estimation of a groupwise additive multiple-index model and its applications. Statist. Sin., 2015, 25: 552-567

[41] YaoYC. Estimating the number of change-points via Schwarz’ criterion. Statist. Probab. Lett., 1988, 6: 181-189

[42] YaoYC, AuST. Least-squares estimation of a step function. Sankhyā Ser A, 1989, 51: 370-381

[43] YinYQ. Detection of the number, locations and magnitudes of jumps. Comm. Statist. Stochastic Models, 1988, 4: 445-455

[44] Yu, M., Chen, X.: Finite sample change point inference and identification for high-dimensional mean vectors. J. R. Statist. Soc B, (to appear) (2021)

[45] ZouCL, YinGS, FengL, WaengZJ. Nonparametric maximum likelihood approach to multiple change-point problems. Ann. Stat., 2014, 42: 970-1002

RIGHTS & PERMISSIONS

School of Mathematical Sciences, University of Science and Technology of China and Springer-Verlag GmbH Germany, part of Springer Nature