asgn1-2010sp

asgn1-2010sp - STA 7249 Generalized Linear Models...

This preview shows pages 1–3. Sign up to view the full content.

STA 7249: Generalized Linear Models Assignment 1 1. This problem concerns the Inverse Gaussian distribution. Let Φ denote the standard normal CDF and consider the function F ( y ) = 0 , y 0 , Φ q λ y - 1 + y μ + e 2 λ/μ Φ - q λ y 1 + y μ , y > 0 . (a) Show that F has density f given by f ( y ) = 0 , y 0 , λ 2 πy 3 1 / 2 exp n - λ ( y - μ ) 2 2 μ 2 y o , y > 0 . (b) Show that the density can be written in exponential dispersion form. Identify: the canonical parameter θ and the dispersion parameter φ (in terms of λ and μ ); the cumulant function, b ( θ ); the canonical link; and the variance function for this model. 2. The idea behind this exercise is to walk you through a way to construct an exponential dispersion family and to derive some basic properties of such a family. The properties that you will derive (and particularly the results of parts 2e through 2g) can be derived directly from the form of the exponential dispersion distribution given in class, so the construction given here is arguably just a curiosity, though it is at least a fairly interesting curiosity. We begin with a review of some basic terminology and results. Let Q represent a probability distribution (i.e., a probability measure) on ( R , R ), where R represents the Borel subsets of R , and let Y be a random variable with distribution Q , i.e., P ( Y A ) = Q ( A ), A R . (Don’t worry too much if you don’t know about the Borel sets: suffice it to say that not all subsets A R are Borel sets, but you’re unlikely to come up with one that isn’t without working pretty hard at it.) The moment generating function (mgf) of Q (or equivalently of Y ) is defined to be M ( t ) = E ( e tY ) = R e ty Q ( dy ). Because e ty > 0, this integral is well-defined for all t , but it may be infinite. Let Θ = { t : M ( t ) < ∞} . Of course M (0) = 1 and thus 0 Θ is assured. (a) Show that Θ is a convex subset of R . The only convex, nonempty subsets of R are intervals (here we count a singleton { t } as an interval through the representation { t } = [ t, t ]), and hence Θ is an interval. It can be shown (see Billingsley, Probability and Measure , 3rd edn) that M ( t ) is dif- ferentiable at all t Θ (the interior of θ ) and that its derivatives can be computed by differentiating across the expectation, i.e., M ( j ) ( t ) = E ( Y j e tY ) for all t Θ . If 1

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
0 Θ , then Y has finite moments of all orders given by E ( X j ) = M ( j ) (0) and its dis- tribution Q is determined (i.e., uniquely identified) by either of its moment generating function or its sequence of moments. Henceforth we will assume that 0 Θ .
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern