Comparison of iterative algorithms for parameter estimation in nonlinear regression

Download

index.pdf

Date

2018

Author

Musluoğlu, Gamze

Metadata

Show full item record

Item Usage Stats

376
views

127
downloads

Nonlinear regression models are more common as compared to linear ones for real life cases e.g. climatology, biology, earthquake engineering, economics etc. However, nonlinear regression models are much more complex to fit and to interpret. Classical parameter estimation methods such as least squares and maximum likelihood can also be adopted to fit the model in nonlinear regression as well, but explicit solutions can not be achieved unlike linear models. At this point, iterative algorithms are utilized to solve the problem numerically. Since there is no extensive study which compiles, classifies and compares the existing methods for nonlinear parameter estimation, the objective of this study is to fill this gap. In our study, we aim to compile the methods which are used for nonlinear parameter estimation purpose and compare them with respect to several criteria such as bias, execution time, number of iterations etc. The comparison will be conducted considering different scenarios which are small vs. large sample sizes, good vs. poor initial values, normal vs. non-normal error terms, simple vs complex models (with respect to number of parameters), and robustness. Both real and simulated data are used in the comparative study.

Subject Keywords

Regression analysis., Nonlinear theories., Computer algorithms.

URI

http://etd.lib.metu.edu.tr/upload/12622525/index.pdf
https://hdl.handle.net/11511/27619

Collections

Graduate School of Natural and Applied Sciences, Thesis

Suggestions

OpenMETU
Core

Structured neural networks for modeling and identification of nonlinear mechanical systems Kılıç, Ergin; Dölen, Melik; Koku, Ahmet Buğra; Department of Mechanical Engineering (2012) Most engineering systems are highly nonlinear in nature and thus one could not develop efficient mathematical models for these systems. Artificial neural networks, which are used in estimation, filtering, identification and control in technical literature, are considered as universal modeling and functional approximation tools. Unfortunately, developing a well trained monolithic type neural network (with many free parameters/weights) is known to be a daunting task since the process of loading a specific pat...
Development of a model updating technique for nonlinear systems Canbaloğlu, Güvenç; Özgüven, Hasan Nevzat; Ünver, Hakkı Özgür; Department of Mechanical Engineering (2015) In structural dynamics, obtaining an accurate numerical model is very crucial. However there are usually discrepancies between calculated dynamic behavior from numerical models and the ones obtained experimentally, and therefore it will be necessary to update the numerical models. In real life applications, structures usually have nonlinearity, and for nonlinear structures, in order to update the numerical model, firstly nonlinearity in the structure can be identified, and then updating procedure may be app...
Fuzzy versus statistical linear regression Kim, KJ; Moskowitz, H; Köksalan, Mustafa Murat (1996-07-19) Statistical linear regression and fuzzy linear regression have been developed from different perspectives, and thus there exist several conceptual and methodological differences between the two approaches. The characteristics of both methods, in terms of basic assumptions, parameter estimation, and application are described and contrasted. Their descriptive and predictive capabilities are also compared via a simulation experiment to identify the conditions under which one outperforms the other. It turns out...
A new contribution to nonlinear robust regression and classification with mars and its applications to data mining for quality control in manufacturing Yerlikaya, Fatma; Weber, Gerhard Wilhelm; Department of Scientific Computing (2008) Multivariate adaptive regression spline (MARS) denotes a modern methodology from statistical learning which is very important in both classification and regression, with an increasing number of applications in many areas of science, economy and technology. MARS is very useful for high dimensional problems and shows a great promise for fitting nonlinear multivariate functions. MARS technique does not impose any particular class of relationship between the predictor variables and outcome variable of interest....
Modeling, inference and optimization of regulatory networks based on time series data Weber, Gerhard Wilhelm; DEFTERLİ, ÖZLEM; ALPARSLAN GÖK, Sırma Zeynep; Kropat, Erik (2011-05-16) In this survey paper, we present advances achieved during the last years in the development and use of OR, in particular, optimization methods in the new gene-environment and eco-finance networks, based on usually finite data series, with an emphasis on uncertainty in them and in the interactions of the model items. Indeed, our networks represent models in the form of time-continuous and time-discrete dynamics, whose unknown parameters we estimate under constraints on complexity and regularization by variou...

Citation Formats

G. Musluoğlu, “Comparison of iterative algorithms for parameter estimation in nonlinear regression,” M.S. - Master of Science, Middle East Technical University, 2018.