Covariance estimation and Graphical Modelling












2














I've started reading on Convariance Matrix estimation through Graphical model in high-dimensional situation. But I have several questions.



Suppose, $X_i overset{iid}{sim} N_p(mu,Sigma)$, $I=1, dots, n$. Define $S=frac{1}{n}sum (X_i-bar{X})(X_i-bar{X})'$. Then almost in every literature, they claim $S$ is really a 'bad estimator' (or performs poorly) in case of large $p$ small $n$. However I understand $S$ is always non-negative definite and symmetric and will be positive definite (w.p. 1) if $ngeq p+1$. Also $hat{Sigma}_{mle}=S$. My question is



1) Why S performs poorly when $p>n$?



2) If $Sigma$ is sparse why do we use different approach to estimate it (methodology involving Graphical representation of $Q=Sigma^{-1}$) ?



BTW I am reading Graphical Models by S. Lauritzen, Gaussian Markov Random Fields by Rue and Held. Please fell free to suggest any other books, materials etc. That you think will help me understand.










share|cite|improve this question





























    2














    I've started reading on Convariance Matrix estimation through Graphical model in high-dimensional situation. But I have several questions.



    Suppose, $X_i overset{iid}{sim} N_p(mu,Sigma)$, $I=1, dots, n$. Define $S=frac{1}{n}sum (X_i-bar{X})(X_i-bar{X})'$. Then almost in every literature, they claim $S$ is really a 'bad estimator' (or performs poorly) in case of large $p$ small $n$. However I understand $S$ is always non-negative definite and symmetric and will be positive definite (w.p. 1) if $ngeq p+1$. Also $hat{Sigma}_{mle}=S$. My question is



    1) Why S performs poorly when $p>n$?



    2) If $Sigma$ is sparse why do we use different approach to estimate it (methodology involving Graphical representation of $Q=Sigma^{-1}$) ?



    BTW I am reading Graphical Models by S. Lauritzen, Gaussian Markov Random Fields by Rue and Held. Please fell free to suggest any other books, materials etc. That you think will help me understand.










    share|cite|improve this question



























      2












      2








      2







      I've started reading on Convariance Matrix estimation through Graphical model in high-dimensional situation. But I have several questions.



      Suppose, $X_i overset{iid}{sim} N_p(mu,Sigma)$, $I=1, dots, n$. Define $S=frac{1}{n}sum (X_i-bar{X})(X_i-bar{X})'$. Then almost in every literature, they claim $S$ is really a 'bad estimator' (or performs poorly) in case of large $p$ small $n$. However I understand $S$ is always non-negative definite and symmetric and will be positive definite (w.p. 1) if $ngeq p+1$. Also $hat{Sigma}_{mle}=S$. My question is



      1) Why S performs poorly when $p>n$?



      2) If $Sigma$ is sparse why do we use different approach to estimate it (methodology involving Graphical representation of $Q=Sigma^{-1}$) ?



      BTW I am reading Graphical Models by S. Lauritzen, Gaussian Markov Random Fields by Rue and Held. Please fell free to suggest any other books, materials etc. That you think will help me understand.










      share|cite|improve this question















      I've started reading on Convariance Matrix estimation through Graphical model in high-dimensional situation. But I have several questions.



      Suppose, $X_i overset{iid}{sim} N_p(mu,Sigma)$, $I=1, dots, n$. Define $S=frac{1}{n}sum (X_i-bar{X})(X_i-bar{X})'$. Then almost in every literature, they claim $S$ is really a 'bad estimator' (or performs poorly) in case of large $p$ small $n$. However I understand $S$ is always non-negative definite and symmetric and will be positive definite (w.p. 1) if $ngeq p+1$. Also $hat{Sigma}_{mle}=S$. My question is



      1) Why S performs poorly when $p>n$?



      2) If $Sigma$ is sparse why do we use different approach to estimate it (methodology involving Graphical representation of $Q=Sigma^{-1}$) ?



      BTW I am reading Graphical Models by S. Lauritzen, Gaussian Markov Random Fields by Rue and Held. Please fell free to suggest any other books, materials etc. That you think will help me understand.







      graph-theory normal-distribution statistical-inference machine-learning estimation-theory






      share|cite|improve this question















      share|cite|improve this question













      share|cite|improve this question




      share|cite|improve this question








      edited Nov 23 at 13:33









      pointguard0

      1,312821




      1,312821










      asked Sep 20 '15 at 22:40









      Bob

      283




      283






















          1 Answer
          1






          active

          oldest

          votes


















          0














          First, I would like to mention that $S$ is always non-negative definite no matter what $p$ and $n$ are. To claim positive definiteness it's not enough to have $n ge p + 1$.



          Now, as for your questions:




          1. When $p > n$ you basically have $frac{p(p+1)}2$ parameters to estimate with $n$ observations only. Imagine, you want to estimate a 3dimensional random vector with only one observation. What would your estimate be? Would that be consistent in any sense?

          2. There are many different methods for sparse covariance matrix estimation, but the reason people don't use estimator $S$ (mentioned in OPs question) is that the sparsity pattern is assumed unknown, in other words, you don't know which elements of matrix $Sigma$ are exactly $0$ and which are not. One of the ways to handle this is to use Graphical models.


          I tried to explain the intuition behind the answers but of course mathematically rigorous explanations also exist.






          share|cite|improve this answer





















            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
            });


            }
            });














            draft saved

            draft discarded


















            StackExchange.ready(
            function () {
            StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f1444217%2fcovariance-estimation-and-graphical-modelling%23new-answer', 'question_page');
            }
            );

            Post as a guest















            Required, but never shown

























            1 Answer
            1






            active

            oldest

            votes








            1 Answer
            1






            active

            oldest

            votes









            active

            oldest

            votes






            active

            oldest

            votes









            0














            First, I would like to mention that $S$ is always non-negative definite no matter what $p$ and $n$ are. To claim positive definiteness it's not enough to have $n ge p + 1$.



            Now, as for your questions:




            1. When $p > n$ you basically have $frac{p(p+1)}2$ parameters to estimate with $n$ observations only. Imagine, you want to estimate a 3dimensional random vector with only one observation. What would your estimate be? Would that be consistent in any sense?

            2. There are many different methods for sparse covariance matrix estimation, but the reason people don't use estimator $S$ (mentioned in OPs question) is that the sparsity pattern is assumed unknown, in other words, you don't know which elements of matrix $Sigma$ are exactly $0$ and which are not. One of the ways to handle this is to use Graphical models.


            I tried to explain the intuition behind the answers but of course mathematically rigorous explanations also exist.






            share|cite|improve this answer


























              0














              First, I would like to mention that $S$ is always non-negative definite no matter what $p$ and $n$ are. To claim positive definiteness it's not enough to have $n ge p + 1$.



              Now, as for your questions:




              1. When $p > n$ you basically have $frac{p(p+1)}2$ parameters to estimate with $n$ observations only. Imagine, you want to estimate a 3dimensional random vector with only one observation. What would your estimate be? Would that be consistent in any sense?

              2. There are many different methods for sparse covariance matrix estimation, but the reason people don't use estimator $S$ (mentioned in OPs question) is that the sparsity pattern is assumed unknown, in other words, you don't know which elements of matrix $Sigma$ are exactly $0$ and which are not. One of the ways to handle this is to use Graphical models.


              I tried to explain the intuition behind the answers but of course mathematically rigorous explanations also exist.






              share|cite|improve this answer
























                0












                0








                0






                First, I would like to mention that $S$ is always non-negative definite no matter what $p$ and $n$ are. To claim positive definiteness it's not enough to have $n ge p + 1$.



                Now, as for your questions:




                1. When $p > n$ you basically have $frac{p(p+1)}2$ parameters to estimate with $n$ observations only. Imagine, you want to estimate a 3dimensional random vector with only one observation. What would your estimate be? Would that be consistent in any sense?

                2. There are many different methods for sparse covariance matrix estimation, but the reason people don't use estimator $S$ (mentioned in OPs question) is that the sparsity pattern is assumed unknown, in other words, you don't know which elements of matrix $Sigma$ are exactly $0$ and which are not. One of the ways to handle this is to use Graphical models.


                I tried to explain the intuition behind the answers but of course mathematically rigorous explanations also exist.






                share|cite|improve this answer












                First, I would like to mention that $S$ is always non-negative definite no matter what $p$ and $n$ are. To claim positive definiteness it's not enough to have $n ge p + 1$.



                Now, as for your questions:




                1. When $p > n$ you basically have $frac{p(p+1)}2$ parameters to estimate with $n$ observations only. Imagine, you want to estimate a 3dimensional random vector with only one observation. What would your estimate be? Would that be consistent in any sense?

                2. There are many different methods for sparse covariance matrix estimation, but the reason people don't use estimator $S$ (mentioned in OPs question) is that the sparsity pattern is assumed unknown, in other words, you don't know which elements of matrix $Sigma$ are exactly $0$ and which are not. One of the ways to handle this is to use Graphical models.


                I tried to explain the intuition behind the answers but of course mathematically rigorous explanations also exist.







                share|cite|improve this answer












                share|cite|improve this answer



                share|cite|improve this answer










                answered Nov 23 at 13:37









                pointguard0

                1,312821




                1,312821






























                    draft saved

                    draft discarded




















































                    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.





                    Some of your past answers have not been well-received, and you're in danger of being blocked from answering.


                    Please pay close attention to the following guidance:


                    • 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.


                    To learn more, see our tips on writing great answers.




                    draft saved


                    draft discarded














                    StackExchange.ready(
                    function () {
                    StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f1444217%2fcovariance-estimation-and-graphical-modelling%23new-answer', 'question_page');
                    }
                    );

                    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







                    Popular posts from this blog

                    Plaza Victoria

                    Puebla de Zaragoza

                    Musa