DTU Studieprojekt - Parameter identification in stochastic differential equations using long-term predictions and estimating functions

Parameter identification in stochastic differential equations using long-term predictions and estimating functions

Udbyder
Vejleder
Sted
København og omegn
This is a theoretical statistics project which aims to develop and/or examine new methods for estimating parameters in stochastic differential equations. This is an on-going research effort in the group, and therre can be a number of different takes on it:

. Theoretical investigation of the approach where we cast the problem as a non-linear mixed effects problem and use numerical methods for optimization of likelihood involving the Laplace approximation, using TMB (Template Model Builder) as the computational engine.

. Investigation of alternatives to Maxium Likelihood estimation, e.g. long-term prediction, in order to increase robustness.

. Case studies and benchmarks. The precise nature of these evolve, but they involve more modelling, data issues, and coding, and in turn less mathematical theory.

Forudsætninger
Time series analysis, stochastic differential equations, likelihood methods in statistics.

Emneord

• Matematik
• Kortlægning og opmåling
• Billedanalyse
• Computerberegning
• Dataanalyse
• Software og programmering
• Geometri
• Matematisk analyse
• Matematisk logik
• Matematisk modellering
• Statistik

Navn
Uffe Høgsbro Thygesen

Stilling
Lektor

Mail
uhth@dtu.dk

Vejleder-info
Kandidatuddannelsen i Matematisk Modellering og Computing
Vejleder
Uffe Høgsbro Thygesen

Type
Kandidatspeciale