Exponential family for Gumbel distribution
up vote
0
down vote
favorite
Consider the Gumbel distributions $(P_vartheta)_{varthetaintheta}=(G(beta,mu))_{(beta,mu)in(0,infty)timesmathbb{R}}$ with distribution functions
$$F_{beta,mu}(x)=e^{-e^{-frac{1}{beta}(x-mu)}}$$
Consider the productmodel $(mathbb{R}^n, mathcal{B}(mathbb{R})^{otimes n}, (P_vartheta^{otimes n})_{varthetaintheta})$
Can $(G(beta,tildemu)^{otimes n})_{betain(0,infty)}$ be written as an exponential family for a given $tildemu$?
Can $(G(tildebeta,mu)^{otimes n})_{muinmathbb{R}}$ be written as an exponential family for a given $tildebeta$?
I found the answer on my own:
So for a given $tildebeta$ my attempt is following:
$$f_mu(x)=frac{1}{tildebeta}e^{-frac{1}{tildebeta}(x-mu)}e^{-e^{-frac{1}{tildebeta}(x-mu)}}$$
Therefore we get
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-sum_{i=1}^ne^{-frac{1}{tildebeta}(x_i-mu)}Big)\
&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-e^{frac{mu}{tildebeta}}sum_{i=1}^ne^{-frac{x_i}{tildebeta}}Big)\
&=expBig(-e^{frac{mu}{tildebeta}}cdotsum_{i=1}^ne^{-frac{x_i}{tildebeta}}+frac{n}{tildebeta}muBig)cdotfrac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig),
end{align}
which is an exponential family.
For given $tildemuinmathbb{R}$ one finds again
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{beta^n}expBig(-frac{1}{beta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{beta}tildemuBig)expBig(-e^{frac{tildemu}{beta}}sum_{i=1}^ne^{-frac{x_i}{beta}}Big)\
end{align}
The critical point is that $$forall iin{1,ldots, n}foralltext{ functions }eta_i,T_i: e^{-frac{x_i}{beta}}ne eta_i(beta)T_i(x_i)$$
Therefore this is not an exponential family, because one would need to to separate the variables $beta$ and $x_i$ in the expression
$$e^{-frac{x_i}{beta}}$$
to write down an exponential family.
probability probability-theory statistics descriptive-statistics
add a comment |
up vote
0
down vote
favorite
Consider the Gumbel distributions $(P_vartheta)_{varthetaintheta}=(G(beta,mu))_{(beta,mu)in(0,infty)timesmathbb{R}}$ with distribution functions
$$F_{beta,mu}(x)=e^{-e^{-frac{1}{beta}(x-mu)}}$$
Consider the productmodel $(mathbb{R}^n, mathcal{B}(mathbb{R})^{otimes n}, (P_vartheta^{otimes n})_{varthetaintheta})$
Can $(G(beta,tildemu)^{otimes n})_{betain(0,infty)}$ be written as an exponential family for a given $tildemu$?
Can $(G(tildebeta,mu)^{otimes n})_{muinmathbb{R}}$ be written as an exponential family for a given $tildebeta$?
I found the answer on my own:
So for a given $tildebeta$ my attempt is following:
$$f_mu(x)=frac{1}{tildebeta}e^{-frac{1}{tildebeta}(x-mu)}e^{-e^{-frac{1}{tildebeta}(x-mu)}}$$
Therefore we get
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-sum_{i=1}^ne^{-frac{1}{tildebeta}(x_i-mu)}Big)\
&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-e^{frac{mu}{tildebeta}}sum_{i=1}^ne^{-frac{x_i}{tildebeta}}Big)\
&=expBig(-e^{frac{mu}{tildebeta}}cdotsum_{i=1}^ne^{-frac{x_i}{tildebeta}}+frac{n}{tildebeta}muBig)cdotfrac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig),
end{align}
which is an exponential family.
For given $tildemuinmathbb{R}$ one finds again
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{beta^n}expBig(-frac{1}{beta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{beta}tildemuBig)expBig(-e^{frac{tildemu}{beta}}sum_{i=1}^ne^{-frac{x_i}{beta}}Big)\
end{align}
The critical point is that $$forall iin{1,ldots, n}foralltext{ functions }eta_i,T_i: e^{-frac{x_i}{beta}}ne eta_i(beta)T_i(x_i)$$
Therefore this is not an exponential family, because one would need to to separate the variables $beta$ and $x_i$ in the expression
$$e^{-frac{x_i}{beta}}$$
to write down an exponential family.
probability probability-theory statistics descriptive-statistics
add a comment |
up vote
0
down vote
favorite
up vote
0
down vote
favorite
Consider the Gumbel distributions $(P_vartheta)_{varthetaintheta}=(G(beta,mu))_{(beta,mu)in(0,infty)timesmathbb{R}}$ with distribution functions
$$F_{beta,mu}(x)=e^{-e^{-frac{1}{beta}(x-mu)}}$$
Consider the productmodel $(mathbb{R}^n, mathcal{B}(mathbb{R})^{otimes n}, (P_vartheta^{otimes n})_{varthetaintheta})$
Can $(G(beta,tildemu)^{otimes n})_{betain(0,infty)}$ be written as an exponential family for a given $tildemu$?
Can $(G(tildebeta,mu)^{otimes n})_{muinmathbb{R}}$ be written as an exponential family for a given $tildebeta$?
I found the answer on my own:
So for a given $tildebeta$ my attempt is following:
$$f_mu(x)=frac{1}{tildebeta}e^{-frac{1}{tildebeta}(x-mu)}e^{-e^{-frac{1}{tildebeta}(x-mu)}}$$
Therefore we get
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-sum_{i=1}^ne^{-frac{1}{tildebeta}(x_i-mu)}Big)\
&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-e^{frac{mu}{tildebeta}}sum_{i=1}^ne^{-frac{x_i}{tildebeta}}Big)\
&=expBig(-e^{frac{mu}{tildebeta}}cdotsum_{i=1}^ne^{-frac{x_i}{tildebeta}}+frac{n}{tildebeta}muBig)cdotfrac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig),
end{align}
which is an exponential family.
For given $tildemuinmathbb{R}$ one finds again
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{beta^n}expBig(-frac{1}{beta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{beta}tildemuBig)expBig(-e^{frac{tildemu}{beta}}sum_{i=1}^ne^{-frac{x_i}{beta}}Big)\
end{align}
The critical point is that $$forall iin{1,ldots, n}foralltext{ functions }eta_i,T_i: e^{-frac{x_i}{beta}}ne eta_i(beta)T_i(x_i)$$
Therefore this is not an exponential family, because one would need to to separate the variables $beta$ and $x_i$ in the expression
$$e^{-frac{x_i}{beta}}$$
to write down an exponential family.
probability probability-theory statistics descriptive-statistics
Consider the Gumbel distributions $(P_vartheta)_{varthetaintheta}=(G(beta,mu))_{(beta,mu)in(0,infty)timesmathbb{R}}$ with distribution functions
$$F_{beta,mu}(x)=e^{-e^{-frac{1}{beta}(x-mu)}}$$
Consider the productmodel $(mathbb{R}^n, mathcal{B}(mathbb{R})^{otimes n}, (P_vartheta^{otimes n})_{varthetaintheta})$
Can $(G(beta,tildemu)^{otimes n})_{betain(0,infty)}$ be written as an exponential family for a given $tildemu$?
Can $(G(tildebeta,mu)^{otimes n})_{muinmathbb{R}}$ be written as an exponential family for a given $tildebeta$?
I found the answer on my own:
So for a given $tildebeta$ my attempt is following:
$$f_mu(x)=frac{1}{tildebeta}e^{-frac{1}{tildebeta}(x-mu)}e^{-e^{-frac{1}{tildebeta}(x-mu)}}$$
Therefore we get
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-sum_{i=1}^ne^{-frac{1}{tildebeta}(x_i-mu)}Big)\
&=frac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{tildebeta}muBig)expBig(-e^{frac{mu}{tildebeta}}sum_{i=1}^ne^{-frac{x_i}{tildebeta}}Big)\
&=expBig(-e^{frac{mu}{tildebeta}}cdotsum_{i=1}^ne^{-frac{x_i}{tildebeta}}+frac{n}{tildebeta}muBig)cdotfrac{1}{tildebeta^n}expBig(-frac{1}{tildebeta}cdotsum_{i=1}^nx_iBig),
end{align}
which is an exponential family.
For given $tildemuinmathbb{R}$ one finds again
begin{align}
f_mu^{otimes n}(x)=prod_{i=1}^n f_mu(x_i)&=frac{1}{beta^n}expBig(-frac{1}{beta}cdotsum_{i=1}^nx_iBig)expBig(frac{n}{beta}tildemuBig)expBig(-e^{frac{tildemu}{beta}}sum_{i=1}^ne^{-frac{x_i}{beta}}Big)\
end{align}
The critical point is that $$forall iin{1,ldots, n}foralltext{ functions }eta_i,T_i: e^{-frac{x_i}{beta}}ne eta_i(beta)T_i(x_i)$$
Therefore this is not an exponential family, because one would need to to separate the variables $beta$ and $x_i$ in the expression
$$e^{-frac{x_i}{beta}}$$
to write down an exponential family.
probability probability-theory statistics descriptive-statistics
probability probability-theory statistics descriptive-statistics
edited Nov 13 at 21:29
asked Nov 13 at 17:17
user408858
19510
19510
add a comment |
add a comment |
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2997010%2fexponential-family-for-gumbel-distribution%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown