Sum of two truncated gaussian
$begingroup$
What is the CDF and the PDF (or approximation) of the sum of two truncated gaussian $X = TN_x(mu_x,sigma_x;a_x,b_x)$ and $Y = TN_y(mu_y,sigma_y;a_y,b_y)$ ? where $TN(mu,sigma;a,b)$ is a truncated normal distribution where a and b are the the lower and upper bounds of the truncation respectively.
probability probability-distributions normal-distribution
$endgroup$
|
show 2 more comments
$begingroup$
What is the CDF and the PDF (or approximation) of the sum of two truncated gaussian $X = TN_x(mu_x,sigma_x;a_x,b_x)$ and $Y = TN_y(mu_y,sigma_y;a_y,b_y)$ ? where $TN(mu,sigma;a,b)$ is a truncated normal distribution where a and b are the the lower and upper bounds of the truncation respectively.
probability probability-distributions normal-distribution
$endgroup$
$begingroup$
Just to clarify, you want to sum distributions or want to know distribution of sum of variables? Because these are different things. Even in not truncated case sum of normal is normal but sum of normal distributions is not normal.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 13:03
$begingroup$
I want to know the distribution of sum of variables. Sorry for the lack of precision.
$endgroup$
– user51213
Jun 12 '14 at 13:50
$begingroup$
Are they independent? Because then we can use the formula:$P_z(z)=E(P_x(z-Y))=E(P_y(z-X)$ where $Z=X+Y$, $P_{x,y,z}$ are CDFs of $x,y,z$ respectively, $E$ is the mean operator.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 15:34
$begingroup$
Are either $a_x$ or $a_y$ positive?
$endgroup$
– wolfies
Jun 12 '14 at 15:36
$begingroup$
Yes, they are independent. What is the form of $P_x(z - Y)$.
$endgroup$
– user51213
Jun 12 '14 at 16:23
|
show 2 more comments
$begingroup$
What is the CDF and the PDF (or approximation) of the sum of two truncated gaussian $X = TN_x(mu_x,sigma_x;a_x,b_x)$ and $Y = TN_y(mu_y,sigma_y;a_y,b_y)$ ? where $TN(mu,sigma;a,b)$ is a truncated normal distribution where a and b are the the lower and upper bounds of the truncation respectively.
probability probability-distributions normal-distribution
$endgroup$
What is the CDF and the PDF (or approximation) of the sum of two truncated gaussian $X = TN_x(mu_x,sigma_x;a_x,b_x)$ and $Y = TN_y(mu_y,sigma_y;a_y,b_y)$ ? where $TN(mu,sigma;a,b)$ is a truncated normal distribution where a and b are the the lower and upper bounds of the truncation respectively.
probability probability-distributions normal-distribution
probability probability-distributions normal-distribution
asked Jun 12 '14 at 12:57
user51213user51213
1113
1113
$begingroup$
Just to clarify, you want to sum distributions or want to know distribution of sum of variables? Because these are different things. Even in not truncated case sum of normal is normal but sum of normal distributions is not normal.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 13:03
$begingroup$
I want to know the distribution of sum of variables. Sorry for the lack of precision.
$endgroup$
– user51213
Jun 12 '14 at 13:50
$begingroup$
Are they independent? Because then we can use the formula:$P_z(z)=E(P_x(z-Y))=E(P_y(z-X)$ where $Z=X+Y$, $P_{x,y,z}$ are CDFs of $x,y,z$ respectively, $E$ is the mean operator.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 15:34
$begingroup$
Are either $a_x$ or $a_y$ positive?
$endgroup$
– wolfies
Jun 12 '14 at 15:36
$begingroup$
Yes, they are independent. What is the form of $P_x(z - Y)$.
$endgroup$
– user51213
Jun 12 '14 at 16:23
|
show 2 more comments
$begingroup$
Just to clarify, you want to sum distributions or want to know distribution of sum of variables? Because these are different things. Even in not truncated case sum of normal is normal but sum of normal distributions is not normal.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 13:03
$begingroup$
I want to know the distribution of sum of variables. Sorry for the lack of precision.
$endgroup$
– user51213
Jun 12 '14 at 13:50
$begingroup$
Are they independent? Because then we can use the formula:$P_z(z)=E(P_x(z-Y))=E(P_y(z-X)$ where $Z=X+Y$, $P_{x,y,z}$ are CDFs of $x,y,z$ respectively, $E$ is the mean operator.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 15:34
$begingroup$
Are either $a_x$ or $a_y$ positive?
$endgroup$
– wolfies
Jun 12 '14 at 15:36
$begingroup$
Yes, they are independent. What is the form of $P_x(z - Y)$.
$endgroup$
– user51213
Jun 12 '14 at 16:23
$begingroup$
Just to clarify, you want to sum distributions or want to know distribution of sum of variables? Because these are different things. Even in not truncated case sum of normal is normal but sum of normal distributions is not normal.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 13:03
$begingroup$
Just to clarify, you want to sum distributions or want to know distribution of sum of variables? Because these are different things. Even in not truncated case sum of normal is normal but sum of normal distributions is not normal.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 13:03
$begingroup$
I want to know the distribution of sum of variables. Sorry for the lack of precision.
$endgroup$
– user51213
Jun 12 '14 at 13:50
$begingroup$
I want to know the distribution of sum of variables. Sorry for the lack of precision.
$endgroup$
– user51213
Jun 12 '14 at 13:50
$begingroup$
Are they independent? Because then we can use the formula:$P_z(z)=E(P_x(z-Y))=E(P_y(z-X)$ where $Z=X+Y$, $P_{x,y,z}$ are CDFs of $x,y,z$ respectively, $E$ is the mean operator.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 15:34
$begingroup$
Are they independent? Because then we can use the formula:$P_z(z)=E(P_x(z-Y))=E(P_y(z-X)$ where $Z=X+Y$, $P_{x,y,z}$ are CDFs of $x,y,z$ respectively, $E$ is the mean operator.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 15:34
$begingroup$
Are either $a_x$ or $a_y$ positive?
$endgroup$
– wolfies
Jun 12 '14 at 15:36
$begingroup$
Are either $a_x$ or $a_y$ positive?
$endgroup$
– wolfies
Jun 12 '14 at 15:36
$begingroup$
Yes, they are independent. What is the form of $P_x(z - Y)$.
$endgroup$
– user51213
Jun 12 '14 at 16:23
$begingroup$
Yes, they are independent. What is the form of $P_x(z - Y)$.
$endgroup$
– user51213
Jun 12 '14 at 16:23
|
show 2 more comments
2 Answers
2
active
oldest
votes
$begingroup$
This is not an easy problem to obtain a closed-form solution to. As always, there are a number of different approaches, but unfortunately many of them seem to yield intractable outcomes. The approach taken here is to proceed manually step-by-step using the Method of Transformations, aided by using a computer algebra system to do the nitty-gritties where that is helpful.
Given: Let:
$X sim N(mu_1, sigma_1^2)$ be doubly truncated (below and above) at $(a_1, b_1)$, where $0<a_1<b_1$, and
$Y sim N(mu_2, sigma_2^2)$ be doubly truncated (below and above) at $(a_2, b_2)$, where $0<a_2<b_2$.
Here is an illustrative plot of the doubly truncated Normal pdf, given different parameter values:
Solution
By virtue of independence, the joint pdf of $(X,Y)$, say $f(x,y)$ is the product of the individual pdf's:
where Erf[.]
denotes the error function.
Step 1: Let $Z=X+Y$ and $W=Y$. Then, using the Method of Transformations, the joint pdf of $(Z,W)$, say $g(z,w)$, is given by:
where:
Transform
is a function from the mathStatica package for Mathematica, which automates the calculation of the transformation and required Jacobian.the transformation $(Z=X+Y, W=Y)$ induces dependency in the domain of support between $Z$ and $W$. In particular, since $X$ is bounded by $(a_1,b_1)$, it follows that $Z=X+Y$ is bounded by $(a_1+W, b_1+W)$. This dependency is captured using the
Boole
statement in the line above which acts as an indicator function. Because the dependency has been captured into pdfg
itself, we can enter the domain of support for pdfg
in standard rectangular fashion as:
Step 2: Given the joint pdf $g(z,w)$ just derived, the marginal pdf of $Z$ is:
which yields the 4-part piecewise solution:
... defined subject to the domain of support: $a_1 + a_2 < Z < b_1 + b_2$.
The solution sol
appears very small on screen: to view it larger, either open the image in a new window, or click here.
The Marginal
computation is far from simple: it takes about 10 minutes of pure computing time to solve on the new R2-D2 Mac Pro.
All done.
Illustrative Plot
Here is a plot of the solution pdf of $Z = X + Y$, for the same parameter values used in the first diagram:
$(mu_1 = 5, sigma_1 = 2, a_1 = 5, b_1 = 8)$
$(mu_2 = 1, sigma_2 = 4, a_2 = 7, b_2 = 9)$
Monte Carlo Check
It is always good idea to check symbolic work with a quick Monte Carlo comparison, just to make sure that no mistakes have crept in. Here is a comparison of the empirical pdf (Monte Carlo simulation of the sum of the doubly truncated Normals) [the squiggly blue curve] plotted together with the exact theoretical pdf (sol
) derived above [ red dashed curve ].
Looks fine!
Notes
- The
Marginal
function used above is also from the mathStatica package for Mathematica. As disclosure, I should add that I am one of the authors of the software used.
$endgroup$
add a comment |
$begingroup$
OK, I believe the variables are independent. Then let $Phi(x,mu,sigma)$ be normal CDF and $rho(x,mu,sigma)$ it's density
$$
rho(x,mu,sigma)=frac{1}{sqrt{2pi}sigma}e^{-frac{(x-mu)^2}{2sigma^2}}
$$
Thus
$$
P_x(x)=(xle a_x)Phi_(a_x,mu_x,sigma_x)+(a_x < x < b_x)int_{a_x}^x rho(x,mu_x,sigma_x) dx+(x>b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
$$
P_y(y)=(yle a_y)Phi_(a_y,mu_y,sigma_y)+(a_y < y < b_y)int_{a_y}^y rho(y,mu_y,sigma_y) dy+(y>b_y)(1-Phi(b_y,mu_y,sigma_y))
$$
So the mean operator $E$ will look as follows:
$$
E(f(X))=Phi(a_x,mu_x,sigma_x)f(a_x)+int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx+f(b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
Now using the formula $P_z(z)=E(P_y(z-X))$ and defining $f(x)=P_y(z-x)$ we can write the following
$$
P_z(z)=Phi(a_x,mu_x,sigma_x)P_y(z-a_x)+int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx+P_y(z-b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
May be going through different inequalities you can simplify this expression, plugging the expression ofr $P_y$ into the last formula.
EDIT
As was noted by wolfies I wrote $P_{x,y}$ for censored variables rather truncated. Thank you for catching me. We can rewrite truncated distribution even simpler:
$$
P_y(y)=frac{1}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}int_{max(y,a_y)}^{min(y,b_y)} rho(y,mu_y,sigma_y) dy =(y>a_y)
frac{Phi(min(y,b_y),mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}
$$
and
$$
E(f(X))=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx
$$
So
$$
P_z(z)=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx
$$
And finally we have very similar formula for density
$$
rho_z(z)=frac{int_{a_x}^{b_x} (a_y<z-x<b_y)rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
It can be simlified further
$$
rho_z(z)=frac{int_{max(a_x,z-b_y)}^{min(b_x,z-a_y)} rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
The last result almost surely can be completed in $Phi$ terms.
$endgroup$
$begingroup$
The cdf of the doubly truncated Normal, evaluated at the lower bound $X = a_x$ is $P_x(a_x) = 0$, ... By contrast, what you have described in your cdf $P_x(x)$ is the cdf of a doubly censored Normal ... not a doubly truncated Normal.
$endgroup$
– wolfies
Jun 12 '14 at 16:51
$begingroup$
Opps.You are right. OK let me edit my answer then
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 16:56
$begingroup$
Thank you. Could you help me calculate $P_z(z)$ ? Thanks in advance.
$endgroup$
– user51213
Jun 13 '14 at 14:44
$begingroup$
Well, I gave finally formula for the density. These are pure exponents of $-(z-x)^2,-x^2)$. They can be recombined to get the final result in terms of $Phi$ .
$endgroup$
– Alexander Vigodner
Jun 13 '14 at 19:53
add a comment |
Your Answer
StackExchange.ifUsing("editor", function () {
return StackExchange.using("mathjaxEditing", function () {
StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix) {
StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
});
});
}, "mathjax-editing");
StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "69"
};
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function() {
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled) {
StackExchange.using("snippets", function() {
createEditor();
});
}
else {
createEditor();
}
});
function createEditor() {
StackExchange.prepareEditor({
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader: {
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
noCode: true, onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});
}
});
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%2f831714%2fsum-of-two-truncated-gaussian%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
2 Answers
2
active
oldest
votes
2 Answers
2
active
oldest
votes
active
oldest
votes
active
oldest
votes
$begingroup$
This is not an easy problem to obtain a closed-form solution to. As always, there are a number of different approaches, but unfortunately many of them seem to yield intractable outcomes. The approach taken here is to proceed manually step-by-step using the Method of Transformations, aided by using a computer algebra system to do the nitty-gritties where that is helpful.
Given: Let:
$X sim N(mu_1, sigma_1^2)$ be doubly truncated (below and above) at $(a_1, b_1)$, where $0<a_1<b_1$, and
$Y sim N(mu_2, sigma_2^2)$ be doubly truncated (below and above) at $(a_2, b_2)$, where $0<a_2<b_2$.
Here is an illustrative plot of the doubly truncated Normal pdf, given different parameter values:
Solution
By virtue of independence, the joint pdf of $(X,Y)$, say $f(x,y)$ is the product of the individual pdf's:
where Erf[.]
denotes the error function.
Step 1: Let $Z=X+Y$ and $W=Y$. Then, using the Method of Transformations, the joint pdf of $(Z,W)$, say $g(z,w)$, is given by:
where:
Transform
is a function from the mathStatica package for Mathematica, which automates the calculation of the transformation and required Jacobian.the transformation $(Z=X+Y, W=Y)$ induces dependency in the domain of support between $Z$ and $W$. In particular, since $X$ is bounded by $(a_1,b_1)$, it follows that $Z=X+Y$ is bounded by $(a_1+W, b_1+W)$. This dependency is captured using the
Boole
statement in the line above which acts as an indicator function. Because the dependency has been captured into pdfg
itself, we can enter the domain of support for pdfg
in standard rectangular fashion as:
Step 2: Given the joint pdf $g(z,w)$ just derived, the marginal pdf of $Z$ is:
which yields the 4-part piecewise solution:
... defined subject to the domain of support: $a_1 + a_2 < Z < b_1 + b_2$.
The solution sol
appears very small on screen: to view it larger, either open the image in a new window, or click here.
The Marginal
computation is far from simple: it takes about 10 minutes of pure computing time to solve on the new R2-D2 Mac Pro.
All done.
Illustrative Plot
Here is a plot of the solution pdf of $Z = X + Y$, for the same parameter values used in the first diagram:
$(mu_1 = 5, sigma_1 = 2, a_1 = 5, b_1 = 8)$
$(mu_2 = 1, sigma_2 = 4, a_2 = 7, b_2 = 9)$
Monte Carlo Check
It is always good idea to check symbolic work with a quick Monte Carlo comparison, just to make sure that no mistakes have crept in. Here is a comparison of the empirical pdf (Monte Carlo simulation of the sum of the doubly truncated Normals) [the squiggly blue curve] plotted together with the exact theoretical pdf (sol
) derived above [ red dashed curve ].
Looks fine!
Notes
- The
Marginal
function used above is also from the mathStatica package for Mathematica. As disclosure, I should add that I am one of the authors of the software used.
$endgroup$
add a comment |
$begingroup$
This is not an easy problem to obtain a closed-form solution to. As always, there are a number of different approaches, but unfortunately many of them seem to yield intractable outcomes. The approach taken here is to proceed manually step-by-step using the Method of Transformations, aided by using a computer algebra system to do the nitty-gritties where that is helpful.
Given: Let:
$X sim N(mu_1, sigma_1^2)$ be doubly truncated (below and above) at $(a_1, b_1)$, where $0<a_1<b_1$, and
$Y sim N(mu_2, sigma_2^2)$ be doubly truncated (below and above) at $(a_2, b_2)$, where $0<a_2<b_2$.
Here is an illustrative plot of the doubly truncated Normal pdf, given different parameter values:
Solution
By virtue of independence, the joint pdf of $(X,Y)$, say $f(x,y)$ is the product of the individual pdf's:
where Erf[.]
denotes the error function.
Step 1: Let $Z=X+Y$ and $W=Y$. Then, using the Method of Transformations, the joint pdf of $(Z,W)$, say $g(z,w)$, is given by:
where:
Transform
is a function from the mathStatica package for Mathematica, which automates the calculation of the transformation and required Jacobian.the transformation $(Z=X+Y, W=Y)$ induces dependency in the domain of support between $Z$ and $W$. In particular, since $X$ is bounded by $(a_1,b_1)$, it follows that $Z=X+Y$ is bounded by $(a_1+W, b_1+W)$. This dependency is captured using the
Boole
statement in the line above which acts as an indicator function. Because the dependency has been captured into pdfg
itself, we can enter the domain of support for pdfg
in standard rectangular fashion as:
Step 2: Given the joint pdf $g(z,w)$ just derived, the marginal pdf of $Z$ is:
which yields the 4-part piecewise solution:
... defined subject to the domain of support: $a_1 + a_2 < Z < b_1 + b_2$.
The solution sol
appears very small on screen: to view it larger, either open the image in a new window, or click here.
The Marginal
computation is far from simple: it takes about 10 minutes of pure computing time to solve on the new R2-D2 Mac Pro.
All done.
Illustrative Plot
Here is a plot of the solution pdf of $Z = X + Y$, for the same parameter values used in the first diagram:
$(mu_1 = 5, sigma_1 = 2, a_1 = 5, b_1 = 8)$
$(mu_2 = 1, sigma_2 = 4, a_2 = 7, b_2 = 9)$
Monte Carlo Check
It is always good idea to check symbolic work with a quick Monte Carlo comparison, just to make sure that no mistakes have crept in. Here is a comparison of the empirical pdf (Monte Carlo simulation of the sum of the doubly truncated Normals) [the squiggly blue curve] plotted together with the exact theoretical pdf (sol
) derived above [ red dashed curve ].
Looks fine!
Notes
- The
Marginal
function used above is also from the mathStatica package for Mathematica. As disclosure, I should add that I am one of the authors of the software used.
$endgroup$
add a comment |
$begingroup$
This is not an easy problem to obtain a closed-form solution to. As always, there are a number of different approaches, but unfortunately many of them seem to yield intractable outcomes. The approach taken here is to proceed manually step-by-step using the Method of Transformations, aided by using a computer algebra system to do the nitty-gritties where that is helpful.
Given: Let:
$X sim N(mu_1, sigma_1^2)$ be doubly truncated (below and above) at $(a_1, b_1)$, where $0<a_1<b_1$, and
$Y sim N(mu_2, sigma_2^2)$ be doubly truncated (below and above) at $(a_2, b_2)$, where $0<a_2<b_2$.
Here is an illustrative plot of the doubly truncated Normal pdf, given different parameter values:
Solution
By virtue of independence, the joint pdf of $(X,Y)$, say $f(x,y)$ is the product of the individual pdf's:
where Erf[.]
denotes the error function.
Step 1: Let $Z=X+Y$ and $W=Y$. Then, using the Method of Transformations, the joint pdf of $(Z,W)$, say $g(z,w)$, is given by:
where:
Transform
is a function from the mathStatica package for Mathematica, which automates the calculation of the transformation and required Jacobian.the transformation $(Z=X+Y, W=Y)$ induces dependency in the domain of support between $Z$ and $W$. In particular, since $X$ is bounded by $(a_1,b_1)$, it follows that $Z=X+Y$ is bounded by $(a_1+W, b_1+W)$. This dependency is captured using the
Boole
statement in the line above which acts as an indicator function. Because the dependency has been captured into pdfg
itself, we can enter the domain of support for pdfg
in standard rectangular fashion as:
Step 2: Given the joint pdf $g(z,w)$ just derived, the marginal pdf of $Z$ is:
which yields the 4-part piecewise solution:
... defined subject to the domain of support: $a_1 + a_2 < Z < b_1 + b_2$.
The solution sol
appears very small on screen: to view it larger, either open the image in a new window, or click here.
The Marginal
computation is far from simple: it takes about 10 minutes of pure computing time to solve on the new R2-D2 Mac Pro.
All done.
Illustrative Plot
Here is a plot of the solution pdf of $Z = X + Y$, for the same parameter values used in the first diagram:
$(mu_1 = 5, sigma_1 = 2, a_1 = 5, b_1 = 8)$
$(mu_2 = 1, sigma_2 = 4, a_2 = 7, b_2 = 9)$
Monte Carlo Check
It is always good idea to check symbolic work with a quick Monte Carlo comparison, just to make sure that no mistakes have crept in. Here is a comparison of the empirical pdf (Monte Carlo simulation of the sum of the doubly truncated Normals) [the squiggly blue curve] plotted together with the exact theoretical pdf (sol
) derived above [ red dashed curve ].
Looks fine!
Notes
- The
Marginal
function used above is also from the mathStatica package for Mathematica. As disclosure, I should add that I am one of the authors of the software used.
$endgroup$
This is not an easy problem to obtain a closed-form solution to. As always, there are a number of different approaches, but unfortunately many of them seem to yield intractable outcomes. The approach taken here is to proceed manually step-by-step using the Method of Transformations, aided by using a computer algebra system to do the nitty-gritties where that is helpful.
Given: Let:
$X sim N(mu_1, sigma_1^2)$ be doubly truncated (below and above) at $(a_1, b_1)$, where $0<a_1<b_1$, and
$Y sim N(mu_2, sigma_2^2)$ be doubly truncated (below and above) at $(a_2, b_2)$, where $0<a_2<b_2$.
Here is an illustrative plot of the doubly truncated Normal pdf, given different parameter values:
Solution
By virtue of independence, the joint pdf of $(X,Y)$, say $f(x,y)$ is the product of the individual pdf's:
where Erf[.]
denotes the error function.
Step 1: Let $Z=X+Y$ and $W=Y$. Then, using the Method of Transformations, the joint pdf of $(Z,W)$, say $g(z,w)$, is given by:
where:
Transform
is a function from the mathStatica package for Mathematica, which automates the calculation of the transformation and required Jacobian.the transformation $(Z=X+Y, W=Y)$ induces dependency in the domain of support between $Z$ and $W$. In particular, since $X$ is bounded by $(a_1,b_1)$, it follows that $Z=X+Y$ is bounded by $(a_1+W, b_1+W)$. This dependency is captured using the
Boole
statement in the line above which acts as an indicator function. Because the dependency has been captured into pdfg
itself, we can enter the domain of support for pdfg
in standard rectangular fashion as:
Step 2: Given the joint pdf $g(z,w)$ just derived, the marginal pdf of $Z$ is:
which yields the 4-part piecewise solution:
... defined subject to the domain of support: $a_1 + a_2 < Z < b_1 + b_2$.
The solution sol
appears very small on screen: to view it larger, either open the image in a new window, or click here.
The Marginal
computation is far from simple: it takes about 10 minutes of pure computing time to solve on the new R2-D2 Mac Pro.
All done.
Illustrative Plot
Here is a plot of the solution pdf of $Z = X + Y$, for the same parameter values used in the first diagram:
$(mu_1 = 5, sigma_1 = 2, a_1 = 5, b_1 = 8)$
$(mu_2 = 1, sigma_2 = 4, a_2 = 7, b_2 = 9)$
Monte Carlo Check
It is always good idea to check symbolic work with a quick Monte Carlo comparison, just to make sure that no mistakes have crept in. Here is a comparison of the empirical pdf (Monte Carlo simulation of the sum of the doubly truncated Normals) [the squiggly blue curve] plotted together with the exact theoretical pdf (sol
) derived above [ red dashed curve ].
Looks fine!
Notes
- The
Marginal
function used above is also from the mathStatica package for Mathematica. As disclosure, I should add that I am one of the authors of the software used.
edited Dec 16 '18 at 8:35
answered Jun 13 '14 at 15:18
wolfieswolfies
4,2092923
4,2092923
add a comment |
add a comment |
$begingroup$
OK, I believe the variables are independent. Then let $Phi(x,mu,sigma)$ be normal CDF and $rho(x,mu,sigma)$ it's density
$$
rho(x,mu,sigma)=frac{1}{sqrt{2pi}sigma}e^{-frac{(x-mu)^2}{2sigma^2}}
$$
Thus
$$
P_x(x)=(xle a_x)Phi_(a_x,mu_x,sigma_x)+(a_x < x < b_x)int_{a_x}^x rho(x,mu_x,sigma_x) dx+(x>b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
$$
P_y(y)=(yle a_y)Phi_(a_y,mu_y,sigma_y)+(a_y < y < b_y)int_{a_y}^y rho(y,mu_y,sigma_y) dy+(y>b_y)(1-Phi(b_y,mu_y,sigma_y))
$$
So the mean operator $E$ will look as follows:
$$
E(f(X))=Phi(a_x,mu_x,sigma_x)f(a_x)+int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx+f(b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
Now using the formula $P_z(z)=E(P_y(z-X))$ and defining $f(x)=P_y(z-x)$ we can write the following
$$
P_z(z)=Phi(a_x,mu_x,sigma_x)P_y(z-a_x)+int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx+P_y(z-b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
May be going through different inequalities you can simplify this expression, plugging the expression ofr $P_y$ into the last formula.
EDIT
As was noted by wolfies I wrote $P_{x,y}$ for censored variables rather truncated. Thank you for catching me. We can rewrite truncated distribution even simpler:
$$
P_y(y)=frac{1}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}int_{max(y,a_y)}^{min(y,b_y)} rho(y,mu_y,sigma_y) dy =(y>a_y)
frac{Phi(min(y,b_y),mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}
$$
and
$$
E(f(X))=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx
$$
So
$$
P_z(z)=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx
$$
And finally we have very similar formula for density
$$
rho_z(z)=frac{int_{a_x}^{b_x} (a_y<z-x<b_y)rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
It can be simlified further
$$
rho_z(z)=frac{int_{max(a_x,z-b_y)}^{min(b_x,z-a_y)} rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
The last result almost surely can be completed in $Phi$ terms.
$endgroup$
$begingroup$
The cdf of the doubly truncated Normal, evaluated at the lower bound $X = a_x$ is $P_x(a_x) = 0$, ... By contrast, what you have described in your cdf $P_x(x)$ is the cdf of a doubly censored Normal ... not a doubly truncated Normal.
$endgroup$
– wolfies
Jun 12 '14 at 16:51
$begingroup$
Opps.You are right. OK let me edit my answer then
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 16:56
$begingroup$
Thank you. Could you help me calculate $P_z(z)$ ? Thanks in advance.
$endgroup$
– user51213
Jun 13 '14 at 14:44
$begingroup$
Well, I gave finally formula for the density. These are pure exponents of $-(z-x)^2,-x^2)$. They can be recombined to get the final result in terms of $Phi$ .
$endgroup$
– Alexander Vigodner
Jun 13 '14 at 19:53
add a comment |
$begingroup$
OK, I believe the variables are independent. Then let $Phi(x,mu,sigma)$ be normal CDF and $rho(x,mu,sigma)$ it's density
$$
rho(x,mu,sigma)=frac{1}{sqrt{2pi}sigma}e^{-frac{(x-mu)^2}{2sigma^2}}
$$
Thus
$$
P_x(x)=(xle a_x)Phi_(a_x,mu_x,sigma_x)+(a_x < x < b_x)int_{a_x}^x rho(x,mu_x,sigma_x) dx+(x>b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
$$
P_y(y)=(yle a_y)Phi_(a_y,mu_y,sigma_y)+(a_y < y < b_y)int_{a_y}^y rho(y,mu_y,sigma_y) dy+(y>b_y)(1-Phi(b_y,mu_y,sigma_y))
$$
So the mean operator $E$ will look as follows:
$$
E(f(X))=Phi(a_x,mu_x,sigma_x)f(a_x)+int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx+f(b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
Now using the formula $P_z(z)=E(P_y(z-X))$ and defining $f(x)=P_y(z-x)$ we can write the following
$$
P_z(z)=Phi(a_x,mu_x,sigma_x)P_y(z-a_x)+int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx+P_y(z-b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
May be going through different inequalities you can simplify this expression, plugging the expression ofr $P_y$ into the last formula.
EDIT
As was noted by wolfies I wrote $P_{x,y}$ for censored variables rather truncated. Thank you for catching me. We can rewrite truncated distribution even simpler:
$$
P_y(y)=frac{1}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}int_{max(y,a_y)}^{min(y,b_y)} rho(y,mu_y,sigma_y) dy =(y>a_y)
frac{Phi(min(y,b_y),mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}
$$
and
$$
E(f(X))=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx
$$
So
$$
P_z(z)=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx
$$
And finally we have very similar formula for density
$$
rho_z(z)=frac{int_{a_x}^{b_x} (a_y<z-x<b_y)rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
It can be simlified further
$$
rho_z(z)=frac{int_{max(a_x,z-b_y)}^{min(b_x,z-a_y)} rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
The last result almost surely can be completed in $Phi$ terms.
$endgroup$
$begingroup$
The cdf of the doubly truncated Normal, evaluated at the lower bound $X = a_x$ is $P_x(a_x) = 0$, ... By contrast, what you have described in your cdf $P_x(x)$ is the cdf of a doubly censored Normal ... not a doubly truncated Normal.
$endgroup$
– wolfies
Jun 12 '14 at 16:51
$begingroup$
Opps.You are right. OK let me edit my answer then
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 16:56
$begingroup$
Thank you. Could you help me calculate $P_z(z)$ ? Thanks in advance.
$endgroup$
– user51213
Jun 13 '14 at 14:44
$begingroup$
Well, I gave finally formula for the density. These are pure exponents of $-(z-x)^2,-x^2)$. They can be recombined to get the final result in terms of $Phi$ .
$endgroup$
– Alexander Vigodner
Jun 13 '14 at 19:53
add a comment |
$begingroup$
OK, I believe the variables are independent. Then let $Phi(x,mu,sigma)$ be normal CDF and $rho(x,mu,sigma)$ it's density
$$
rho(x,mu,sigma)=frac{1}{sqrt{2pi}sigma}e^{-frac{(x-mu)^2}{2sigma^2}}
$$
Thus
$$
P_x(x)=(xle a_x)Phi_(a_x,mu_x,sigma_x)+(a_x < x < b_x)int_{a_x}^x rho(x,mu_x,sigma_x) dx+(x>b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
$$
P_y(y)=(yle a_y)Phi_(a_y,mu_y,sigma_y)+(a_y < y < b_y)int_{a_y}^y rho(y,mu_y,sigma_y) dy+(y>b_y)(1-Phi(b_y,mu_y,sigma_y))
$$
So the mean operator $E$ will look as follows:
$$
E(f(X))=Phi(a_x,mu_x,sigma_x)f(a_x)+int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx+f(b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
Now using the formula $P_z(z)=E(P_y(z-X))$ and defining $f(x)=P_y(z-x)$ we can write the following
$$
P_z(z)=Phi(a_x,mu_x,sigma_x)P_y(z-a_x)+int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx+P_y(z-b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
May be going through different inequalities you can simplify this expression, plugging the expression ofr $P_y$ into the last formula.
EDIT
As was noted by wolfies I wrote $P_{x,y}$ for censored variables rather truncated. Thank you for catching me. We can rewrite truncated distribution even simpler:
$$
P_y(y)=frac{1}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}int_{max(y,a_y)}^{min(y,b_y)} rho(y,mu_y,sigma_y) dy =(y>a_y)
frac{Phi(min(y,b_y),mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}
$$
and
$$
E(f(X))=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx
$$
So
$$
P_z(z)=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx
$$
And finally we have very similar formula for density
$$
rho_z(z)=frac{int_{a_x}^{b_x} (a_y<z-x<b_y)rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
It can be simlified further
$$
rho_z(z)=frac{int_{max(a_x,z-b_y)}^{min(b_x,z-a_y)} rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
The last result almost surely can be completed in $Phi$ terms.
$endgroup$
OK, I believe the variables are independent. Then let $Phi(x,mu,sigma)$ be normal CDF and $rho(x,mu,sigma)$ it's density
$$
rho(x,mu,sigma)=frac{1}{sqrt{2pi}sigma}e^{-frac{(x-mu)^2}{2sigma^2}}
$$
Thus
$$
P_x(x)=(xle a_x)Phi_(a_x,mu_x,sigma_x)+(a_x < x < b_x)int_{a_x}^x rho(x,mu_x,sigma_x) dx+(x>b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
$$
P_y(y)=(yle a_y)Phi_(a_y,mu_y,sigma_y)+(a_y < y < b_y)int_{a_y}^y rho(y,mu_y,sigma_y) dy+(y>b_y)(1-Phi(b_y,mu_y,sigma_y))
$$
So the mean operator $E$ will look as follows:
$$
E(f(X))=Phi(a_x,mu_x,sigma_x)f(a_x)+int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx+f(b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
Now using the formula $P_z(z)=E(P_y(z-X))$ and defining $f(x)=P_y(z-x)$ we can write the following
$$
P_z(z)=Phi(a_x,mu_x,sigma_x)P_y(z-a_x)+int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx+P_y(z-b_x)(1-Phi(b_x,mu_x,sigma_x))
$$
May be going through different inequalities you can simplify this expression, plugging the expression ofr $P_y$ into the last formula.
EDIT
As was noted by wolfies I wrote $P_{x,y}$ for censored variables rather truncated. Thank you for catching me. We can rewrite truncated distribution even simpler:
$$
P_y(y)=frac{1}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}int_{max(y,a_y)}^{min(y,b_y)} rho(y,mu_y,sigma_y) dy =(y>a_y)
frac{Phi(min(y,b_y),mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}{Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y)}
$$
and
$$
E(f(X))=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} f(x)rho(x,mu_x,sigma_x) dx
$$
So
$$
P_z(z)=frac{1}{Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x)}int_{a_x}^{b_x} P_y(z-x)rho(x,mu_x,sigma_x) dx
$$
And finally we have very similar formula for density
$$
rho_z(z)=frac{int_{a_x}^{b_x} (a_y<z-x<b_y)rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
It can be simlified further
$$
rho_z(z)=frac{int_{max(a_x,z-b_y)}^{min(b_x,z-a_y)} rho(z-x,mu_y,sigma_y)rho(x,mu_x,sigma_x) dx}{(Phi(b_x,mu_x,sigma_x)-Phi(a_x,mu_x,sigma_x))(Phi(b_y,mu_y,sigma_y)-Phi(a_y,mu_y,sigma_y))}
$$
The last result almost surely can be completed in $Phi$ terms.
edited Jun 12 '14 at 18:17
answered Jun 12 '14 at 16:20
Alexander VigodnerAlexander Vigodner
2,0391713
2,0391713
$begingroup$
The cdf of the doubly truncated Normal, evaluated at the lower bound $X = a_x$ is $P_x(a_x) = 0$, ... By contrast, what you have described in your cdf $P_x(x)$ is the cdf of a doubly censored Normal ... not a doubly truncated Normal.
$endgroup$
– wolfies
Jun 12 '14 at 16:51
$begingroup$
Opps.You are right. OK let me edit my answer then
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 16:56
$begingroup$
Thank you. Could you help me calculate $P_z(z)$ ? Thanks in advance.
$endgroup$
– user51213
Jun 13 '14 at 14:44
$begingroup$
Well, I gave finally formula for the density. These are pure exponents of $-(z-x)^2,-x^2)$. They can be recombined to get the final result in terms of $Phi$ .
$endgroup$
– Alexander Vigodner
Jun 13 '14 at 19:53
add a comment |
$begingroup$
The cdf of the doubly truncated Normal, evaluated at the lower bound $X = a_x$ is $P_x(a_x) = 0$, ... By contrast, what you have described in your cdf $P_x(x)$ is the cdf of a doubly censored Normal ... not a doubly truncated Normal.
$endgroup$
– wolfies
Jun 12 '14 at 16:51
$begingroup$
Opps.You are right. OK let me edit my answer then
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 16:56
$begingroup$
Thank you. Could you help me calculate $P_z(z)$ ? Thanks in advance.
$endgroup$
– user51213
Jun 13 '14 at 14:44
$begingroup$
Well, I gave finally formula for the density. These are pure exponents of $-(z-x)^2,-x^2)$. They can be recombined to get the final result in terms of $Phi$ .
$endgroup$
– Alexander Vigodner
Jun 13 '14 at 19:53
$begingroup$
The cdf of the doubly truncated Normal, evaluated at the lower bound $X = a_x$ is $P_x(a_x) = 0$, ... By contrast, what you have described in your cdf $P_x(x)$ is the cdf of a doubly censored Normal ... not a doubly truncated Normal.
$endgroup$
– wolfies
Jun 12 '14 at 16:51
$begingroup$
The cdf of the doubly truncated Normal, evaluated at the lower bound $X = a_x$ is $P_x(a_x) = 0$, ... By contrast, what you have described in your cdf $P_x(x)$ is the cdf of a doubly censored Normal ... not a doubly truncated Normal.
$endgroup$
– wolfies
Jun 12 '14 at 16:51
$begingroup$
Opps.You are right. OK let me edit my answer then
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 16:56
$begingroup$
Opps.You are right. OK let me edit my answer then
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 16:56
$begingroup$
Thank you. Could you help me calculate $P_z(z)$ ? Thanks in advance.
$endgroup$
– user51213
Jun 13 '14 at 14:44
$begingroup$
Thank you. Could you help me calculate $P_z(z)$ ? Thanks in advance.
$endgroup$
– user51213
Jun 13 '14 at 14:44
$begingroup$
Well, I gave finally formula for the density. These are pure exponents of $-(z-x)^2,-x^2)$. They can be recombined to get the final result in terms of $Phi$ .
$endgroup$
– Alexander Vigodner
Jun 13 '14 at 19:53
$begingroup$
Well, I gave finally formula for the density. These are pure exponents of $-(z-x)^2,-x^2)$. They can be recombined to get the final result in terms of $Phi$ .
$endgroup$
– Alexander Vigodner
Jun 13 '14 at 19:53
add a comment |
Thanks for contributing an answer to Mathematics Stack Exchange!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
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%2f831714%2fsum-of-two-truncated-gaussian%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
$begingroup$
Just to clarify, you want to sum distributions or want to know distribution of sum of variables? Because these are different things. Even in not truncated case sum of normal is normal but sum of normal distributions is not normal.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 13:03
$begingroup$
I want to know the distribution of sum of variables. Sorry for the lack of precision.
$endgroup$
– user51213
Jun 12 '14 at 13:50
$begingroup$
Are they independent? Because then we can use the formula:$P_z(z)=E(P_x(z-Y))=E(P_y(z-X)$ where $Z=X+Y$, $P_{x,y,z}$ are CDFs of $x,y,z$ respectively, $E$ is the mean operator.
$endgroup$
– Alexander Vigodner
Jun 12 '14 at 15:34
$begingroup$
Are either $a_x$ or $a_y$ positive?
$endgroup$
– wolfies
Jun 12 '14 at 15:36
$begingroup$
Yes, they are independent. What is the form of $P_x(z - Y)$.
$endgroup$
– user51213
Jun 12 '14 at 16:23