Abstract
In this paper, we investigate a linear regression time series model of possibly outlier-afflicted observations and autocorrelated random deviations. This colored noise is represented by a covariance-stationary autoregressive (AR) process, in which the independent error components follow a scaled (Student’s) t-distribution. This error model allows for the stochastic modeling of multiple outliers and for an adaptive robust maximum likelihood (ML) estimation of the unknown regression and AR coefficients, the scale parameter, and the degree of freedom of the t-distribution. This approach is meant to be an extension of known estimators, which tend to focus only on the regression model, or on the AR error model, or on normally distributed errors. For the purpose of ML estimation, we derive an expectation conditional maximization either algorithm, which leads to an easy-to-implement version of iteratively reweighted least squares. The estimation performance of the algorithm is evaluated via Monte Carlo simulations for a Fourier as well as a spline model in connection with AR colored noise models of different orders and with three different sampling distributions generating the white noise components. We apply the algorithm to a vibration dataset recorded by a high-accuracy, single-axis accelerometer, focusing on the evaluation of the estimated AR colored noise model.
Similar content being viewed by others
Notes
We denote unknown parameters by Greek letters, random variables by calligraphic letters, and constants by Roman letters. Thus, we distinguish between a random variable (e.g., \(\mathcal {E}_t\)) and its realization \((e_t)\). Furthermore, matrices and vectors are represented by bold letters.
One ADXL (ADXL345 chip from Analog Devices company with Arduino UNO Board), one NAMS (BNO055 chip with Arduino UNO board and Nine-Axis-Motion-Shield), and one IMU (Bosch BNO055-Chip Tinkerforge-Board).
References
Abramowitz M, Stegun IA (1972) Handbook of mathematical functions. Dover Publications, New York
Alkhatib H, Schuh WD (2007) Integration of the Monte Carlo covariance estimation strategy into tailored solution procedures for large-scale least squares problems. J Geod 81(1):53–66. doi:10.1007/s00190-006-0034-z
Alkhatib H, Kargoll B, Neumann I, Kreinovich V (2017) Normalization-invariant fuzzy logic operations explain empirical success of student distributions in describing measurement uncertainty. Technical Report UTEP-CS-17-55, Department of Computer Science, The University of Texas at El Paso. To appear in: proceedings of the 2017 annual conference of the North American fuzzy information processing society NAFIPS’2017, Cancun, Mexico, 16–18 Oct 2017
Amiri-Simkooei AR, Tiberius CCJM, Teunissen PJG (2007) Assessment of noise in GPS coordinate time series: methodology and results. J Geophys Res. doi:10.1029/2006JB004913
Arato M (1961) On the sufficient statistics for stationary Gaussian random processes. Theor Prob Appl 6(2):199–201
Baarda W (1968) A testing procedure for use in geodetic networks. Publications on Geodesy (New Series), vol 2, 5th edn. Netherlands Geodetic Commission, Delft
Bera AK, Jarque CM (1982) Model specification tests: a simultaneous approach. J Econom 20(1):59–82. doi:10.1016/0304-4076(82)90103-8
Bernardo JM (1976) Algorithm AS 103 psi (digamma function) computation. J R Stat Soc C 25:315–317
Bilmes JA (1998) A gentle tutorial of the EM algorithm and its application to parameter estimation for Gaussian mixture and hidden Markov models. In: International computer science institute, Berkeley. http://www.icsi.berkeley.edu/ftp/global/pub/techreports/1997/tr-97-021.pdf. Accessed 17 Jul 2017
Box GEP, Andersen SL (1955) Permutation theory in the derivation of robust criteria and the study of departures from assumption. J R Stat Soc B 17(1):1–34
Brockmann JM (2015) On high performance computing in geodesy: applications in global gravity field determination. Deutsche Geodätische Kommission, Series C (Dissertations), no 761, Munich
Brockmann JM, Kargoll B (2012) Uncertainty assessment of some data-adaptive M-estimators. In: Sneeuw N, Novák P, Crespi M, Sansò F (eds) Proceedings of the VII Hotine–Marussi symposium on mathematical geodesy. International association of geodesy symposia, vol 137. Springer, Berlin, pp 87–92. doi:10.1007/978-3-642-22078-4_13
Brockmann JM, Kargoll B, Krasbutter I, Schuh W-D, Wermuth M (2010) GOCE data analysis: from calibrated measurements to the global earth gravity field. In: Flechtner F, Mandea M, Gruber T, Rothacher M, Wickert J, Güntner A (eds) System earth via geodetic-geophysical space techniques. Advanced technologies in earth sciences. Springer, Berlin, pp 213–229. doi:10.1007/978-3-642-10228-8_17
Bronstein IN, Semendjajew KA (1991) Taschenbuch der Mathematik, 25th edn. Teubner, Stuttgart
Bureick J, Alkhatib H, Neumann I (2016) Robust spatial approximation of laser scanner point clouds by means of free-form curve approaches in deformation analysis. J Appl Geod 10(1):27–35. doi:10.1515/jag-2015-0020
Cai J, Sneeuw N (2014) Stochastic modeling of GOCE gravitational tensor invariants. In: Observation of the system earth from space—CHAMP, GRACE, GOCE and future missions, GEOTECHNOLOGIEN science report No. 20. Springer, Berlin, pp 115–121. doi:10.1007/978-3-642-32135-1_15
Chang XW, Guo Y (2005) Huber’s M-estimation in relative GPS positioning: computational aspects. J Geod 79:351–362. doi:10.1007/s00190-005-0473-y
Christman M (2015) Calibration of industrial accelerometers. IMI division of PCB Piezotronics, Depew, New York. http://www.pcb.com/contentstore/mktgcontent/IMI_Downloads/calibration-of-industrial-accelerometers.pdf. Accessed 17 Jul 2017
Christmas J, Everson R (2011) Robust autoregression: student-t innovations using variational Bayes. IEEE Trans Signal Process 59:48–57. doi:10.1109/TSP.2010.2080271
Dempster AP, Laird NM, Rubin DB (1977) Maximum likelihood from incomplete data via the EM algorithm. J R Stat Soc B 39:1–38
Dempster AP, Laird NM, Rubin DB (1980) Iteratively reweighted least squares for linear regression when errors are normal/independent distributed. In: Krishnaiah PR (ed) Multivariate analysis-V. North-Holland, Amsterdam, pp 35–57
Dodge Y, Jureckova J (1987) Adaptive combination of least squares and least absolute derivations estimators. In: Dodge Y (ed) Statistical data analysis based on the L1-norm and related methods. North Holland, Amsterdam, pp 275–284
Dodge Y, Jureckova J (1988) Adaptive combination of M-estimator and L1-estimator. In: Dodge Y, Fedorov VV, Wynn HP (eds) Optimal design and analysis of experiments. North Holland, Amsterdam, pp 167–176
Duchnowski R (2009) Geodetic application of R-estimation: levelling network examples. Tech Sci 12:135–144
Duchnowski R (2013) Hodges–Lehmann estimates in deformation analyses. J Geod 87(10–12):873–884. doi:10.1007/s00190-013-0651-2
Durbin J (1960) The fitting of time series models. Rev Inst Int Stat 28(3):233–244
Förstner W, Wrobel BP (2016) Photogrammetric computer vision. Springer, New York
Fraser DAS (1976) Necessary analysis and adaptive inference. J Am Stat Assoc 71(353):99–110
Geweke J (1993) Bayesian treatment of the independent Student-t linear model. J Appl Econom 8(S1):19–40. doi:10.1002/jae.3950080504
Götzelmann M, Keller W, Reubelt T (2006) Gross error compensation for gravity field analysis based on kinematic orbit data. J Geod 80:184–198. doi:10.1007/s00190-006-0061-9
Guo J, Ou J, Wang H (2010) Robust estimation for correlated observations: two local sensitivity-based downweighting strategies. J Geod 84:243–250. doi:10.1007/s00190-009-0361-y
Gupta MR, Chen Y (2011) Theory and use of the EM algorithm. Found Trends Signal Process 4(3):223–296. doi:10.1561/2000000034
Hamilton JD (1994) Time series analysis. Princeton University Press, New Jersey
Hampel F (2001) Robust statistics: a brief introduction and overview. In: Carosio A, Kutterer H (eds) First international symposium on robust statistics and fuzzy techniques in geodesy and GIS. Institute of Geodesy and Photogrammetry, ETH Zurich, Zurich
Hekimoglu S, Berber M (2003) Effectiveness of robust methods in heterogeneous linear models. J Geod 76:706–713. doi:10.1007/s00190-002-0289-y
Helmert FR (1907) Die Ausgleichungsrechnung nach der Methode der kleinsten Quadrate, 2nd edn. Teubner, Leipzig
Heyde CC, Leonenko NN (2005) Student processes. Adv Appl Probab 37:342–365. doi:10.1239/aap/1118858629
Hogg RV (1974) Adaptive robust procedures: a partial review and some suggestions for future applications and theory. J Am Stat Assoc 69:909–923
Huber PJ (1964) Robust estimation of a location parameter. Ann Math Stat 35(1):73–101
Junhuan P (2005) The asymptotic variance-covariance matrix, Baarda test and the reliability of \(L_1\)-norm estimates. J Geod 78:668–682. doi:10.1007/s00190-004-0433-y
Kargoll B (2005) Comparison of some robust parameter estimation techniques for outlier analysis applied to simulated GOCE mission data. In: Jekeli C, Bastos L, Fernandes J (eds) Gravity, geoid and space missions. International association of geodesy symposia, vol 129. Springer, Berlin, pp 77–82. doi:10.1007/3-540-26932-0_14
Kargoll B (2012) On the theory and application of model misspecification tests in geodesy. Deutsche Geodätische Kommission, Series C (Dissertations), no 674, Munich
Kemkes E (2016) Suitability testing of micro-electro-mechanical-systems (MEMS) for integration into a geosensor network for the automated monitoring of the swing structures. Master thesis. Leibniz Universität Hannover, Geodetic Institute
Klees R, Ditmar P, Broersen P (2003) How to handle colored observation noise in large least-squares problems. J Geod 76(11):629–640. doi:10.1007/s00190-002-0291-4
Klees R, Ditmar P, Kusche J (2004) Numerical techniques for large least-squares problems with applications to GOCE. In: Sansó F (ed) Proceedings of the V Hotine-Marussi symposium on mathematical geodesy, international association of geodesy symposia, vol 127. Springer, Berlin, pp 12–21. doi:10.1007/978-3-662-10735-5_3
Knight NL, Wang J, Rizos C (2010) Generalised measures of reliability for multiple outliers. J Geod 84(10):625–635. doi:10.1007/s00190-010-0392-4
Koch KR (1999) Parameter estimation and hypothesis testing in linear models. Springer, Berlin
Koch KR (2007) Introduction to Bayesian statistics, 2nd edn. Springer, Berlin
Koch KR (2013a) Robust estimation by expectation maximization algorithm. J Geod 87:107–116. doi:10.1007/s00190-012-0582-3
Koch KR (2013) Comparison of two robust estimations by expectation maximization algorithms with Huber’s method and outlier tests. J Appl Geod 7:115–123. doi:10.1515/jag-2013-0050
Koch KR (2012) Robust estimation by expectation maximization algorithm. J Geod 87(2):107–116. doi:10.1007/s00190-012-0582-3
Koch KR (2014) Outlier detection for the nonlinear Gauss Helmert model with variance components by the expectation maximization algorithm. J Appl Geod 8(3):185–194. doi:10.1515/jag-2014-0004
Koch KR (2017) Expectation maximization algorithm and its minimal detectable outliers. Stud Geophys Geod 61:1–18. doi:10.1007/s11200-016-0617-y
Koch KR, Kargoll B (2013) Expectation-maximization algorithm for the variance-inflation model by applying the t distribution. J Appl Geod 7:217–225. doi:10.1515/jag-2013-0007
Koch KR, Kargoll B (2015) Outlier detection by the EM algorithm for laser scanning in rectangular and polar systems. J Appl Geod 9(3):162–173. doi:10.1515/jag-2015-0004
Koch KR, Schmidt M (1994) Deterministische und stochastische Signale. Dümmler
Kok JJ (1984) On data snooping and multiple outlier testing. NOAA Technical Report NOS NGS 30. US Department of Commerce, National Geodetic Survey, Rockville/Maryland
Krarup T, Juhl J, Kubik K (1980) Götterdämmerung over least squares adjustment. In: Proceedings of the 14th congress ISP Hamburg, international archives of photogrammetry, XXIII, B3. Commission III, pp 369–378
Krasbutter I, Brockmann JM, Kargoll B, Schuh W-D, Goiginger H, Pail R (2011) Refinement of the stochastic model of GOCE scientific data in a long time series. In: Ouwehand L (ed) Proceedings of the 4th international GOCE user workshop. ESA Publication SP-696
Krasbutter I, Brockmann JM, Kargoll B, Schuh WD (2014) Adjustment of digital filters for decorrelation of GOCE SGG data. In: Flechtner F, Sneeuw N, Schuh WD (eds) Observation of the system earth from space: CHAMP, GRACE, GOCE and future missions. GEOTECHNOLOGIEN Science Report no 20. Springer, Berlin, pp 109–114. doi:10.1007/978-3-642-32135-1_14
Krasbutter I, Kargoll B, Schuh W-D (2015) Magic square of real spectral and time series analysis with an application to moving average processes. In: Kutterer H, Seitz F, Alkhatib H, Schmidt M (eds) The 1st international workshop on the quality of geodetic observation and monitoring systems (QuGOMS’11) International association of geodesy symposia, vol 140. Springer, Berlin, pp 9–14. doi:10.1007/978-3-319-10828-5_2
Kubik K (1982) An error theory for the Danish method. In: Proceedings of the symposium of the International Society of Photogrammetry, Commission III, Helsinki
Kusolitsch N (2014) Mass-und Wahrscheinlichkeitstheorie. Springer, Berlin
Lange KL, Little RJA, Taylor JMG (1989) Robust statistical modeling using the t distribution. J Am Stat Assoc 84:881–896. doi:10.2307/2290063
Lehmann R (2013) \(3\sigma \)-rule for outlier detection from the viewpoint of geodetic adjustment. J Surv Eng 139(4):157–165. doi:10.1061/(ASCE)SU.1943-5428.0000112
Lehmann R (2015) Observation error model selection by information criteria vs. normality testing. Stud Geophys Geod 59(4):489–504. doi:10.1007/s11200-015-0725-0
Li L (2011) Separablity of deformations and measurement noises of GPS time series with modified Kalman filter for landslide monitoring in real-time. Ph.D. thesis, Institute of Geodesy and Geoinformation, Universität Bonn
Lindenberger J (1993) Laser-Profilmessungen zur topographischen Geländeaufnahme. Deutsche Geodätische Kommission, Series C (Dissertations), no 400, Munich
Little RJA (1988) Robust estimation of the mean and covariance matrix from data with missing values. Appl Stat 37(1):23–38
Little RJA, Rubin DB (2002) Statistical analysis with missing data, 2nd edn. Wiley, Hoboken
Liu CH, Rubin DB (1994) The ECME algorithm: a simple extension of EM and ECM with faster monotone convergence. Biometrika 81:633–648. doi:10.1093/biomet/81.4.633
Liu CH, Rubin DB (1995) ML estimation of the t distribution using EM and its extensions, ECM and ECME. Stat Sin 5:19–39
Liu CH (1997) ML estimation of the multivariate t distribution and the EM algorithm. J Multivar Anal 63:296–312. doi:10.1006/jmva.1997.1703
Luo X, Mayer M, Heck B (2012) Analysing time series of GNSS residuals by means of AR(I)MA processes. In: Sneeuw N, Novàk P, Crespi M, Sansó F (eds) Proceedings of the VII Hotine–Marussi symposium on mathematical geodesy International association of geodesy symposia, vol 137. Springer, Berlin, pp 129–134. doi:10.1007/978-3-642-22078-4_19
Luxen M, Brunn A (2003) Parameterschätzung aus unvollständigen Beobachtungsdaten mittels des EM-Algorithmus. Z Geod Geoinform Landmanag 128:71–78
Marshall J (2002) L1-norm pre-analysis measures for geodetic networks. J Geod 76:334–344. doi:10.1007/s00190-002-0254-9
Marx C (2013) On resistant Lp-norm estimation by means of iteratively reweighted least squares. J Appl Geod 7(1):1–10. doi:10.1515/jag-2012-0042
McDonald JB (1989) Partially adaptive estimation of ARMA time series models. Int J Forecast 5:217–230. doi:10.1016/0169-2070(89)90089-7
McLachlan GJ, Krishnan T (2008) The EM algorithm and extensions, 2nd edn. Wiley, Hoboken
Meng X-L, Rubin DB (1993) Maximum likelihood estimation via the ECM algorithm: a general framework. Biometrika 80:267–278
Metaxoglu K, Smith A (2007) Maximum likelihood estimation of VARMA models using a state-space EM algorithm. J Time Ser Anal 28(5):666–685. doi:10.1111/j.1467-9892.2007.00529.x
Nadarajah S (2009) The product t density distribution arising from the product of two Student’s t PDFs. Stat Pap 50:605–615. doi:10.1007/s00362-007-0088-x
Nassar S, El-Sheimy N (2005) Accuracy improvement of stochastic modeling of inertial sensor errors. Z Geod Geoinform Landmanag 130:146–155
Nassar S, Schwarz KP, El-Sheimy N, Noureldin A (2004) Modeling inertial sensor errors using autoregressive (AR) models. Navigation 51:259–268. doi:10.1002/j.2161-4296.2004.tb00357.x
Niedzielski T, Kosek W (2010) An application of low-order ARMA and GARCH models for sea level fluctuations. Artif Satell 45(1):27–39. doi:10.2478/v10018-010-0003-x
Niedzielski T, Kosek W (2012) Prediction analysis of UT1-UTC time series by combination of the least-squares and multivariate autoregressive method. In: Sneeuw N, Novàk P, Crespi M, Sansó F (eds) Proceedings of the VII Hotine–Marussi symposium on mathematical geodesy. International association of geodesy symposia, vol 137. Springer, Berlin, pp 153–157. doi:10.1007/978-3-642-22078-4_23
Nowel K, Kamiński W (2014) Robust estimation of deformation from observation differences for free control networks. J Geod 88(8):749–764. doi:10.1007/s00190-014-0719-7
Orlov A (1991) How often are the observations normal? Ind Lab 57(7):770–772
Pail R, Bruinsma S, Migliaccio F, Förste C, Goiginger H, Schuh WD, Höck E, Reguzzoni M, Brockmann JM, Abrikosov O, Veicherts M, Fecher T, Mayrhofer R, Krasbutter I, Sansò F, Tscherning CG (2011) First GOCE gravity field models derived by three different approaches. J Geod 85:819–843. doi:10.1007/s00190-011-0467-x
Park M, Gao Y (2008) Error and performance analysis of MEMS-based inertial sensors with a low-cost GPS receiver. Sensors 8(4):2240–2261. doi:10.3390/s8042240
Parzen E (1999) Stochastic processes. Society for industrial and applied mathematics (SIAM). Holden Day, Oakland
Peng J (2009) Jointly robust estimation of unknown parameters and variance components based on expectation-maximization algorithm. J Surv Eng 135:1–9. doi:10.1061/(ASCE)0733-9453(2009)
Pope AJ (1976) The statistics of residuals and the detection of outliers. NOAA Technical Report NOS65 NGS1, US Department of Commerce, National Geodetic Survey, Rockville, Maryland
Porat B (1994) Digital processing of random signals. Dover Publications, Mineola
Priestley MB (1981) Spectral analysis and time series. Academic Press, Cambridge
Resnick SI (2007) Heavy-tail phenomena: probabilistic and statistical modeling. Springer, New York
Sanubari J (1999) Robust spectral estimation based on ARMA model excited by a t-distribution process. In: Proceedings of the IEEE-EURASIP workshop on nonlinear signal and image processing (NSIP’99), vol 2, pp 607–611
Sanubari J, Tokuda K, Onoda M (1992) Spectral estimation based on AR-model, excited by t-distribution process. Proc IEEE Int Conf Acoust Speech Signal Process 5:521–524. doi:10.1109/ICASSP.1992.226568
Schall J, Eicker A, Kusche J (2014) The ITG-Goce02 gravity field model from GOCE orbit and gradiometer data based on the short arc approach. J Geod 88(4):403–409. doi:10.1007/s00190-014-0691-2
Schlittgen R, Streitberg BHJ (2001) Zeitreihenanalyse, 9th edn. R. Oldenbourg Verlag, Munich
Schön S, Kermarrec G, Kargoll B, Neumann I, Kosheleva O, Kreinovich V (2017) Why student distributions? Why Matern’s covariance model? A symmetry-based explanation. Technical Report UTEP-CS-17-65, Department of Computer Science, The University of Texas at El Paso
Schuh WD (2003) The processing of band-limited measurements; filtering techniques in the least squares context and in the presence of data gaps. Space Sci Rev 108(1):67–78. doi:10.1023/A:1026121814042
Schuh WD, Brockmann JM, Kargoll B, Krasbutter I, Pail R (2010) Refinement of the stochastic model of GOCE scientific data and its effect on the in-situ gravity field solution. In: Lacoste-Francis H (ed) Proceedings of the ESA Living planet symposium. ESA publication SP-686, ESA/ESTEC. ISBN (Online) 978-9292212506, ISSN 1609-042X
Schuh WD, Krasbutter I, Kargoll B (2014) Korrelierte Messung: was nun? In: Neuner H (ed) Zeitabhängige Messgrössen - Ihre Daten haben (Mehr-)Wert
Siemes C (2013) Digital Filtering Algorithms for Decorrelation within Large Least Squares Problems. Deutsche Geodätische Kommission, Series C (Dissertations), no 707, Munich
Student (Gosset WS), (1908) The probable error of a mean. Biometrika 6:1–25. doi:10.1093/biomet/6.1.1
Tiku ML, Wong W-K, Vaughan DC, Bian G (2000) Time series models in non-normal situations: symmetric innovations. J Time Ser Anal 21:571–596. doi:10.1111/1467-9892.00199
Tukey JW (1960) A survey of sampling from contaminated distributions. In: Olkin I (ed) Contributions to probability and statistics: essays in honor of Harold Hotelling. Stanford University Press, Stanford, pp 448–485
Wang K, Xiong S, Li Y (2012) Modeling with noises for inertial sensors. In: Position location and navigation symposium (PLANS) 2012. IEEE/ION, pp 625–632. doi:10.1109/PLANS.2012.6236937
Wiśniewski Z (2009) Estimation of parameters in a split functional model of geodetic observations (\(M_{\rm split}\) estimation). J Geod 83(2):105–120. doi:10.1007/s00190-008-0241-x
Wiśniewski Z (2014) \(M\)-estimation with probabilistic models of geodetic observations. J Geod 88(10):941–957. doi:10.1007/s00190-014-0735-7
Xu PL (1988) Application of time series methods in prediction of displacements on large dams. J Wuhan Tech Univ Surv Mapp 13(3):23–31
Xu PL (1993) Consequences of constant parameters and confidence intervals of robust estimation. Boll Geod Sci Affini 52:231–249
Xu PL (2002) A hybrid global optimization method: the one-dimensional case. J Comput Appl Math 147(2):301–314. doi:10.1016/S0377-0427(02)00438-7
Xu PL (2003) A hybrid global optimization method: the multi-dimensional case. J Comput Appl Math 155(2):423–446. doi:10.1016/S0377-0427(02)00878-6
Xu PL (2005) Sign-constrained robust least squares, subjective breakdown point and the effect of weights of observations on robustness. J Geod 79:146–159. doi:10.1007/s00190-005-0454-1
Yang Y, Song L, Xu T (2002) Robust estimator for correlated observations based on bifactor equivalent weights. J Geod 76:353–358. doi:10.1007/s00190-002-0256-7
Zhong D (1997) Robust estimation and optimal selection of polynomial parameters for the interpolation of GPS geoid heights. J Geod 71(9):552–561. doi:10.1007/s001900050123
Acknowledgements
We thank the editors and the reviewers for their constructive comments and valuable suggestions, which helped to improve this paper. The presented application of the PCB Piezotronics accelerometer within the vibration analysis experiment was performed as a part of the collaborative project “Spatio-temporal monitoring of bridge structures using low cost sensors” with ALLSAT GmbH, which is funded by the German Federal Ministry for Economic Affairs and Energy (BMWi) and the Central Innovation Programme for SMEs (ZIM Kooperationsprojekt, ZF4081803DB6). In addition, the authors would like to acknowledge the Institute of Concrete Construction (Leibniz Universität Hannover) for providing the shaker table and the reference accelerometer used within this experiment.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix
A Used distributions and useful properties
This section reviews the main distributions as well as corresponding densities and expected values used throughout the paper. This compilation is partly based on Koch (1999), but also offers other results that are less well known. Throughout this section, \(\mathcal {X}\) denotes any real-valued random variable (r.v.).
1.1 A. 1 The univariate, scaled (Student’s) t-distribution
If \(\mathcal {X}\) follows a univariate, scaled Student’s t-distribution with \(\nu >0\) degrees of freedom, location parameter \(\mu \) and scale parameter \(\sigma > 0\), symbolically \(\mathcal {X} \sim t_{\nu }(\mu ,\sigma )\), then the density of \(\mathcal {X}\) is given by
Here \(\mathrm {\Gamma }(.)\) denotes the gamma function (cf. Abramowitz and Stegun 1972, p. 255). If \(\nu > 1\), then the expectation of \(\mathcal {X}\) is defined and equals \(\mu \); if \(\nu > 2\), then the variance of \(\mathcal {X}\) is finite and equals \(\frac{\nu }{\nu -2} \cdot \sigma ^2\) (cf. Lange et al. 1989). Setting \(\sigma = 1\) leads to the (unscaled) univariate Student’s t-distribution introduced by Student (Gosset WS) (1908).
1.2 A.2 The univariate gamma distribution
If \(\mathcal {X}\) follows a univariate gamma distribution with parameters \(a>0\) and \(b>0\), symbolically \(\mathcal {X} \sim G(a,b)\), then the density of \(\mathcal {X}\) is given by
The chi-squared distribution (say, with \(\nu \) degrees of freedom) is a special case of the gamma distribution in the sense that \(\chi ^2_{\nu } = G \left( \frac{1}{2}, \frac{\nu }{2} \right) \).
The moment generating function of G(a, b) is given by
which allows for the convenient determination of the expectation of \(\mathcal {X}\) via the relationship \(E_{\mathcal {X}} = M'_{G(a,b)}(0)\), where \(M'(.)\) denotes the derivative of the moment generating function. It is easily verified that differentiation of (69) with respect to t and evaluation of that derivative for \(t=0\) yields the value \(\frac{a}{b}\) for the expectation of \(\mathcal {X}\).
Another property we exploit in this paper is the following scaling property. If \(s \ne 0\), then
Using this property and the above-mentioned relationship between the gamma and the chi-squared distribution, we see that
Next, we show how the expected value of the natural logarithm of a gamma-distributed random variable can be obtained, that is,
Using (68) and the substitution \(z = bx\) (with \(x > 0\), thus \(z > 0\)), we obtain
The first integral takes the value \(\mathrm {\Gamma }'(a)\) (cf. p. 331, for the derivatives of the gamma function Bronstein and Semendjajew 1991), and the second integral takes the value \(\mathrm {\Gamma }(a)\) by definition of the gamma function. Thus, we obtain
where \(\psi (.)\) denotes the digamma (or psi) function (cf. Abramowitz and Stegun 1972, p. 258f.). To evaluate the digamma function we used the standard MATLAB function call psi(0,x); an algorithm for this purpose is given in Bernardo (1976).
B Factorization of the likelihood function
We obtain for the argument of the exponential function in (31) the \(N = (p+1)(p+2)/2\) terms
the further N terms
as well as the N terms
and finally, the single term \(\frac{\nu }{2} \varvec{1}_n^T \bar{\varvec{w}}\). All of the \(3N + 1\) occurring scalar products are functions of the augmented data \(\varvec{y} = (\varvec{u},\varvec{w})\) and do not involve any parameters. Therefore, they may be viewed as data-dependent functions, as defined by (33)–(36).
C Derivation of the M-step
1.1 C.1 Regarding the CM-step for \(\varvec{\xi }\)
Concerning the jth functional parameter \(\xi _j\) (\(j = 1, \ldots , m\)), we seek the solution of
or equivalently of
Writing all m of these equations jointly, we find (43).
1.2 C.2 CM-step for \(\varvec{\alpha }\)
Recalling (42) and (2), with respect to the h-th autoregressive coefficient \(\alpha _h\) (\(h \in \{ 1,\ldots ,p \}\)), we obtain for the first-order condition
Writing all p equations in matrix notation, we thus have
Denoting the solution of this equation system by \(\varvec{\xi } = \varvec{\xi }^{(k+1)}\) and \(\varvec{\alpha } = \varvec{\alpha }^{(k+1)}\), and using furthermore the notations (48)–(49), we find the normal equations
leading to (50).
1.3 C.3 CM-step for \(\sigma ^2\)
Solving the first-order condition
for \(\sigma ^2\) by using the already available estimates \(\varvec{\xi }^{(k+1)}\), \(\varvec{\alpha }^{(k+1)}\), \(\varvec{e}^{(k+1)}\) and (51), we obtain (52).
1.4 C.4 CM-step for \(\nu \)
The first-order condition with respect to \(\nu \) reads
Multiplying with 2 / n and using the definition of the digamma function, a zero \(\nu ^{(k+1)}\) of this equation satisfies (53).
Next, we derive the CME-Step (55). Rewriting the log-likelihood function (14) in the form
we obtain for the first-order condition
Substituting here \(\varvec{\xi }^{(k+1)}\), \(\varvec{\alpha }^{(k+1)}\), \((\sigma ^2)^{(k+1)}\), \(\varvec{e}^{(k+1)}\) from (48), and finally \(\varvec{u}^{(k+1)}\) from (51), it follows from the preceding equation that
Using notation (54) and multiplying the preceding equation with 2 / n then yields finally (55).
D Data for the Monte Carlo simulation
Rights and permissions
About this article
Cite this article
Kargoll, B., Omidalizarandi, M., Loth, I. et al. An iteratively reweighted least-squares approach to adaptive robust adjustment of parameters in linear regression models with autoregressive and t-distributed deviations. J Geod 92, 271–297 (2018). https://doi.org/10.1007/s00190-017-1062-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00190-017-1062-6