Asymptotic Inference for a Linear Stochastic Differential Equation with Time Delay
read more
Citations
Statistical Inference for Stochastic Processes.
Parametric estimation for linear stochastic differential equations driven by fractional Brownian motion
Delay Estimation for Some Stationary Diffusion-type Processes
Parameter estimation for the stochastic SIS epidemic model
Parameter estimation for the stochastic SIS epidemic model
References
Introduction to Functional Differential Equations
Limit Theorems for Stochastic Processes
Theory of martingales
Weak Limit Theorems for Stochastic Integrals and Stochastic Differential Equations
Asymptotics in statistics : some basic concepts
Related Papers (5)
Frequently Asked Questions (6)
Q2. What is the simplest proof of the XT Y T L?
T Y T L fW fWwhere fW t t is a standard Wiener process independent of fW Since XT s W T s s a by It o s formula the authors also haveXT Y T L fX fWwhere fX s a Z sfW t dfW t Moreover the convergence implies the joint functional convergence of XT Y T together with their quadratic co variations see Jacod and Shiryaev Theorem VI
Q3. Where is the family P R locally asymptotically mixed normal?
In Case M the family P R is locally asymptotically mixed normal atVT IT d V Iwhere V The authord The authorZ The authorand Z is an independent of The authorand N The authordistributed vector
Q4. What is the simplest way to calculate the residuals in v?
Then by using Cauchy s residual theorem the authors getx t XRe uRes t lim wi Z u iw u iw t d twhere t e th t Here the authors have used that j t j tends to zero uniformly on u iw v iw and on u iw v iw if jwj Now observe that either v if b v a or v i for some if b v a The explicit calculation of the residuals in v in the rst case and in and in the second case yields the form given in Lemma
Q5. What is the proof of the Fubini inequality?
The proof is trivial and therefore omittedLemma Assume Y t Y t and Z t t are adapted continuous processes Y t Y t Y t t and W t t is a standard Wiener process Moreover let T and T be normalizing functions such thatT Z T Y t dt T and T Z T Z t dt Tare bounded in probability andT Z T Y t dt PThenTZ T Y t dW tZ T Y t dW tPTZ T Y t dt Z T Y t dtPT TZ T Y t Z t dt Z T Y t Z t dtPLet Y t t be a process having the representation Sometimes the rst term in the right hand side of is small in the sense of Lemma i e it can be chosen as Y t The next lemma shows that then the second term in the right hand side of is also small in the same senseLemma Putz t Z y t s X s ds t tThen Z T t z t dtZ X s ds Z T y t dtFor the proof use Fubini s theorem and the Cauchy Schwartz inequality In Lemmas and Corollary the authors assume that Y Y Y are continuous processes having the representation with functions y y y respectivelyLemma Assume that y y t t is a square integrable function ThenT Z T Y t dt PT Z T Y t dt P Zy t dtProof According to Lemmas and it is su'cient to prove the assertion for X s The authors introduce the stationary process Z t R t y t s dW s t where W is independently of W s s extended to as a Wiener process Obviously the authors haveT E Z T t Z t Y t dt T Z T t Z t y s ds dtApplying the law of large numbers to the Gaussian stationary process Z which is ergodic the authors getT Z T Z t dt P EZ T Z T Z t dt P EZNow the claim follows fromCorollary
Q6. what is the maximum likelihood estimator of the true parameter?
The maximum likelihood estimator T of the true parameter a b is given byT arg max R T The authorT V TwhereT V TI T RV TZ T X t dX t Z T X t dX tandI TBB Z T X t dt Z T X t X t dtZ TX t X t dtZ T X t dt CCA Choose an arbitrary nonsingular matrix T and introduce a new parameter R byTThen T T Twhere T is de ned by T arg maxRT The authorT VTwithT VTITVT TZ T X t dW t Z T X t dW tandIT TI T TFrom Section the authors know that under appropriate choice of T the authors haveVT IT d V Ior Vu n