Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I know that Gauss-Markov process expected value is always zero but in normal distribution, mean can vary and Gauss-Markov process is a subtype of Gaussian process.

The reason behind this question is my homework. I have to find $E[y(x)]$ and $R_y[t_1,t_2]$ assuming $x(t)$ is Gaussian process. The transformation is linear: $y(t)=t^{2}x(t)+e^{-x(t)}$. If I assume $E[x(t)]=0$ the solutions are very easy, for example mean: $E[y(t)]=E[t^{2}x(t)]+E[\lambda e^{-\lambda x(t)}]=t^{2}E[x(t)]+\frac{1}{\lambda}=0+1=1$ I am unsure if I understand Gaussian process behaviour properly.

share|improve this question
    
A Gaussian process need not have zero mean; indeed, $E[X(t)]$ can be a function of $t$. Also, even if $E[X(t)] = 0$ for all $t$, it is not true that $$E[y(t)] = E[t^2X(t) + e^{-X(t)}] = 0 + 1 = 1.$$ Note that $E[e^{-X(t)}]$ will give you the value of the moment generating function of the Gaussian random variable $X(t)$ at $-1$. –  Dilip Sarwate Feb 13 '12 at 15:59
    
Thanks, especially for pointing out moment generating function. –  Maciej Kucia Feb 13 '12 at 17:29
    
I would understand the term "Gaussian process" to mean a process $X$ in which for every finite set of times $t_1,\ldots,t_n$ and every set of real coefficients $c_1,\ldots,c_n$, the linear combination $c_1 X(t_1)+\cdots+c_n X(t_n)$ has a Gaussian distribution. That doesn't require zero mean. But maybe a definition posited in some particular book doesn't do it that way. –  Michael Hardy Feb 13 '12 at 17:35

1 Answer 1

up vote 1 down vote accepted

This is a matter of definition. Some textbooks or authors will take "Gaussian process" to include the requirement that the process have mean zero; others do not. You should check the definition in your book, or ask your instructor.

share|improve this answer
    
Thanks for help. I will ask my instructor, unfortunately he has a winter break right now. For now the best I can do is to find answers assuming $E[x(t)]=m(t)$ and after that checking solution for $m(t)=0$. –  Maciej Kucia Feb 13 '12 at 17:33

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.