CINXE.COM

<!DOCTYPE html> <html lang="en"> <head> <meta charset="utf-8"> <meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no"> <link rel="alternate" href="https://www.alliedacademies.org/" hreflang="en" /> <link rel="canonical" href="https://www.alliedacademies.org/articles/more-on-the-orthogonal-complement-functions-11697.html"> <link rel="shortcut icon" href="https://www.alliedacademies.org/images/favicon.ico"/>  <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/jqueryui/1.12.1/jquery-ui.css" /> <link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.1.1/css/bootstrap.min.css"> <link href="https://fonts.googleapis.com/css?family=Open+Sans:300,300i,400,400i,600,600i,700,700i,800,800i" rel="stylesheet"> <link href="https://fonts.googleapis.com/css?family=Catamaran:100,200,300,400,500,600,700,800,900" rel="stylesheet"> <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/animate.css/3.5.2/animate.min.css" /> <link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.7.1/css/all.css"> <link rel="stylesheet" href="https://www.alliedacademies.org/css/metisMenu.min.css"> <link rel="stylesheet" href="https://www.alliedacademies.org/css/dzsparallaxer.css"> <link rel="stylesheet" href="https://www.alliedacademies.org/assets/css/icons.css"> <link rel="stylesheet" href="https://www.alliedacademies.org/assets/css/author.css"> <link rel="stylesheet" href="https://www.alliedacademies.org/css/global.css"> <link rel="stylesheet" href="https://www.alliedacademies.org/css/styles.css">   <script async src="https://www.googletagmanager.com/gtag/js?id=G-Q6TCJGJW4D"></script> <script> window.dataLayer = window.dataLayer || []; function gtag(){dataLayer.push(arguments);} gtag('js', new Date()); gtag('config', 'G-Q6TCJGJW4D'); </script> <title> More on the orthogonal complement functions </title> <meta name="keywords" content="Covariance structure analysis, Distribution free tests, Implicit function theorem, QR factorization"> <meta name="description" content="Continuous orthogonal complement functions have had an interesting history in covariance structure analysis. They were used in a seminal paper by Browne in his development of a d.. "> <meta name="citation_publisher" content="Allied Academies"/> <meta name="citation_journal_title" content="Journal of Applied Mathematics and Statistical Applications"> <meta name="citation_title" content="More on the orthogonal complement functions"> <meta name="citation_author" content="Robert Jennrich"/> <meta name="citation_author" content="Albert Satorr"/> <meta name="citation_year" content="2019"> <meta name="citation_volume" content="2"> <meta name="citation_issue" content="1"> <meta name="citation_issn" content="Open Access"> <meta name="citation_publication_date" content=""/> <meta name="citation_abstract" content="Continuous orthogonal complement functions have had an interesting history in covariance structure analysis. They were used in a seminal paper by Browne in his development of a distribution-free goodness of fit test for an arbitrary covariance structure. The proof of his main result Proposition 4 used a locally continuous orthogonal complement function, but because he failed to show such functions existed his proof was incomplete. In spite of the fact that his test had been used extensively, this problem was not noticed until 2013 when Jennrich and Satorra pointed out that his proof was incomplete and completed it by showing that locally continuous orthogonal complement functions exist. This was done using the implicit function theorem. A problem with the implicit function approach is that it does not give a formula for the locally continuous function produced. This problem was potentially solved by Browne and Shapiro who gave a very simple formula F(X) for an orthogonal complement of X. Unfortunately, they failed to prove that their function actually produced orthogonal complements. We will prove that given a p譹 matrix X0 with full column rank q"> <meta name="citation_fulltext_html_url" content="https://www.alliedacademies.org/articles/more-on-the-orthogonal-complement-functions-11697.html"> <meta name="citation_pdf_url" content="https://www.alliedacademies.org/articles/more-on-the-orthogonal-complement-functions.pdf"> <meta name="citation_abstract_html_url" content="https://www.alliedacademies.org/abstract/more-on-the-orthogonal-complement-functions-11697.html"> <script type="text/javascript"> function openimage(theURL, winName, features) { window.open(theURL, winName, features); } </script> </head> <body class="border-top-3 border-danger"> <svg class="d-none"> <defs> <path id="tabshape" d="M80,60C34,53.5,64.417,0,0,0v60H80z"/> </defs> </svg> <header> <nav class="navbar navbar-expand-lg navbar-light"> <div class="container"> <a class="navbar-brand w-xs-50" href="https://www.alliedacademies.org/" title="Allied Academies"> <img src="https://www.alliedacademies.org/images/allied-academies-logo.png" alt="" class="max-height-60 img-fluid"> </a> <button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#mainNav" aria-controls="mainNav" aria-expanded="false" aria-label="Toggle navigation"> <span class="navbar-toggler-icon"></span> </button> <div class="collapse navbar-collapse justify-content-end" id="mainNav"> <div class="navbar-nav"> <a class="nav-item nav-link active" href="https://www.alliedacademies.org/" title="Allied Academies">Home</a> <a class="nav-item nav-link" href="https://www.alliedacademies.org/journals.php" title="A-Z Journals">A-Z Journals</a> <li class="nav-item dropdown"> <a class="nav-link dropdown-toggle" href="#" id="About" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false" title="About"> About </a> <div class="dropdown-menu animated fadeInUp" aria-labelledby="About"> <a class="dropdown-item" href="https://www.alliedacademies.org/#" title="FAQs">FAQs</a> <a class="dropdown-item" href="https://www.alliedacademies.org/history.php" title="History">History</a> </div> </li> <li class="nav-item dropdown"> <a class="nav-link dropdown-toggle" href="#" id="GuidelinesPolicies" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false" title="Guidelines and Policies"> Guidelines & Policies </a> <div class="dropdown-menu animated fadeInUp" aria-labelledby="GuidelinesPolicies"> <a class="dropdown-item" href="https://www.alliedacademies.org/publication-policies-and-procedures.php" title="Publication Policies and Procedures">Publication Policies and Procedures</a> <a class="dropdown-item" href="https://www.alliedacademies.org/publication-guidelines.php" title="Publication Guidelines">Publication Guidelines</a> <a class="dropdown-item" href="https://www.alliedacademies.org/journal-submission-instructions.php" title="Journal Submission Instructions">Journal Submission Instructions</a> <a class="dropdown-item" href="https://www.alliedacademies.org/code-of-conduct.php" title="Code of Conduct">Code of Conduct</a> <a class="dropdown-item" href="https://www.alliedacademies.org/ethics.php" title="Ethics">Ethics</a> </div> </li> <a class="nav-item nav-link" href="https://www.alliedacademies.org/contact.php" title="Contact">Contact</a> <div id="google_translate_element" align="center" style="width:180px;"></div> <script> function googleTranslateElementInit() { new google.translate.TranslateElement( { pageLanguage: 'en' }, 'google_translate_element' ); } </script> <script src="https://translate.google.com/translate_a/element.js?cb=googleTranslateElementInit"></script> </div> </div> </div> </nav> <section class="bg-grey-200"> <div class="container py-4"> <div class="row align-items-center justify-content-between"> <div class="col-12 col-sm-12"> <h2 class="text-danger mb-0 font-size-26 fweight-500">Journal of Applied Mathematics and Statistical Applications</h2> </div>  </div> </div> </section>  <div class="container"> <div class="media alert alert-primary mt-3"> <i class="fas fa-hand-point-right fa-2x align-self-center mr-3 animated infinite bounce slow blue-500"></i> <div class="media-body"> All submissions of the EM system will be redirected to <strong>Online Manuscript Submission System</strong>. Authors are requested to submit articles directly to <a href="https://www.scholarscentral.org/submissions/applied-mathematics-statistical-applications.html" target="_blank" class="text-warning" title="Online Manuscript Submission System"><strong>Online Manuscript Submission System</strong></a> of respective journal. </div> </div> </div>  </header>  <div class="phonering-alo-phone phonering-alo-green phonering-alo-show bg-primary p-2 rounded-right" id="feedback"> <div class="phonering-alo-ph-circle"></div> Reach Us <img src="https://www.imedpub.com/images/telephone-white.svg" alt="" width="20" class="ml-2"> <img src="https://www.imedpub.com/images/whatsapp.svg" alt="" width="20" class="ml-2"> +44-7360-538437</div> <section class="py-4"> <div class="container"> <div class="row"> <main class="col-xs-12 col-sm-9"> <div class="card border-5 rounded-0"> <article class="card-body full-text"> <p class="grey-600 mb-1"><i class="fa fa-caret-right" aria-hidden="true"></i> Short Communication - Journal of Applied Mathematics and Statistical Applications (2019) Volume 2, Issue 1</p> <h1 class="text-danger font-size-26 fweight-400">More on the orthogonal complement functions</h1> <strong>Robert Jennrich<sup><a href="#a1">1</a><a href="#corr">*</a></sup> and Albert Satorr<sup><a href="#a2">2</a></sup></strong> <p><sup>1</sup><a name="a1" id="a1"></a>Department of Mathematics, University of California, Los Angeles, USA</p> <p><sup>2</sup><a name="a2" id="a2"></a>Department of Mathematics, Universitat Pompeu Fabra, Barcelona, Spain</p> <dl class="dl-horizontal"> <dt>*Corresponding Author:</dt> <dd>Robert Jennrich<a name="corr" id="corr"></a><br> Department of Mathematics,University of California<br> Los Angeles, USA<br> <strong>Tel:</strong> (310) 825-2207<br> <strong>E-mail:</strong> rij@stat.ucla.edu</dd> </dl> <p><strong>Accepted Date:</strong> February 23, 2019</p> <p><strong>Citation: </strong>Jennrich R, Satorr A. More on the orthogonal complement functions. J Appl Math Statist Appl. 2019;2(1):47-50.<br> <div> <strong>Visit for more related articles at </strong> <a href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/volume-selector.php" title="Journal of Applied Mathematics and Statistical Applications"> Journal of Applied Mathematics and Statistical Applications </a> </div>  <h2 class="post-title">Abstract</h2> Continuous orthogonal complement functions have had an interesting history in covariance structure analysis. They were used in a seminal paper by Browne in his development of a distribution-free goodness of fit test for an arbitrary covariance structure. The proof of his main result Proposition 4 used a locally continuous orthogonal complement function, but because he failed to show such functions existed his proof was incomplete. In spite of the fact that his test had been used extensively, this problem was not noticed until 2013 when Jennrich and Satorra pointed out that his proof was incomplete and completed it by showing that locally continuous orthogonal complement functions exist. This was done using the implicit function theorem. A problem with the implicit function approach is that it does not give a formula for the locally continuous function produced. This problem was potentially solved by Browne and Shapiro who gave a very simple formula F(X) for an orthogonal complement of X. Unfortunately, they failed to prove that their function actually produced orthogonal complements. We will prove that given a p譹 matrix X0 with full column rank q<p, then F(X) is an orthogonal complement of X for all X sufficiently close to X0. Much more is true. We will give a necessary and sufficient condition on X under which F(X) is an orthogonal complement of X. We will also show the X for which F(X) is not an orthogonal complement of X is a subset with Lebesgue measure zero in Rpxq. <h2 class="post-title">Keywords</h2> <p>Covariance structure analysis, Distribution free tests, Implicit function theorem, QR factorization.</p> <h2 class="post-title">Introduction</h2> <p>The orthogonal complement of a <em>p×q</em> matrix X with <em>q<p</em> and full column rank is a <em>p</em>×(<em>p-q</em>) matrix <em>Y </em>such that [X,Y] is invertible. Note that <em>Y</em> must have full column rank. Jennrich and Satorra in Theorem 1 show how to compute an orthogonal complement <em>Y</em> of an arbitrary <em>p×q</em> matrix <em>X</em> with full column rank <em>q<p</em> using the long form of a QR factorization. Unfortunately, <em>Y</em> is not a continuous function of X [<a href="#1" title="1">1</a>].</p> <p>In a seminal paper Browne showed how to test the goodness of fit of an arbitrary covariance structure [<a href="#2" title="2">2</a>]. The proof of his main result Proposition 4 used a locally continuous orthogonal complement function, but because he failed to show such functions existed his proof was incomplete. In spite of the fact that his test had been used extensively this problem was not noticed until 2013 when Jennrich and Satorra pointed out that his proof was incomplete and completed it by showing that locally continuous orthogonal complement functions exist [<a href="#1" title="1">1</a>]. This was done using the implicit function theorem. A problem with the implicit function approach is that it does not give a formula for the function produced and Jennrich and Satorra conjectured their function could not be expressed by an explicit formula.</p> <p>Browne and Shapiro say that this conjecture is incorrect by presenting an explicit formula for a locally continuous orthogonal complement function [<a href="#3" title="3">3</a>]. Browne and Shapiro in</p> <p>slightly different notation state the following: Let <em>X</em><sub>0</sub> be any <em>p×q</em> matrix with full column rank <em>q</em><<em>p</em> and X<sup>c</sup><sub>0</sub> be any orthogonal complement of <em>X</em><sub>0</sub>. Consider the following matrix valued function:</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e001.png" alt="equation" /></p> <p>Let N be any neighbourhood of <em>X</em><sub>0</sub> such that <em>X</em> has full column rank for all <em>X</em> ϵ <em>N</em>. Browne and Shapiro say that on <em>N, F(X)</em> is a well-defined continuous function and claim <em>F(X)</em> is an orthogonal complement of <em>X</em> for all <em>X</em> in <em>N</em>. They, however, fail to show this. In particular they fail to show <em>F(X)</em> has full column rank. As a consequence their claim is only a conjecture. We will begin by proving their conjecture.</p> <p><strong>Proof of the Browne-Shapiro conjecture</strong></p> <p><strong>Lemma 1:</strong> Let <em>X</em><sub>0</sub> be a <em>p</em>×<em>q</em> matrix of full column rank <em>q</em><<em>p</em>. There is a neighbourhood <em>N</em> of <em>X</em><sub>0</sub> that contains only full column rank matrices.</p> <p><strong>Proof:</strong> Let X be an arbitrary p×q matrix and let ɡ(<em>X</em>) =det(<em>X</em>'<em>X</em> ) . Since ɡ(<em>X</em> ) is continuous there is <em>δ</em>>0 such that</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e002.png" alt="equation" /></p> <p>It follows from the last inequality that ɡ( <em>X</em> ) ≠ 0 and hence that det(<em>X</em>'<em>X</em> ) ≠ 0 and hence that <em>X</em> has full column rank for all <em>X</em> in the neighbourhood</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e003.png" alt="equation" /></p> <p><strong>Lemma 2:</strong> Let <em>X</em><sub>0</sub> be a <em>p</em>×<em>q</em> matrix with full column rank <em>q</em><<em>p</em> and let <em>X</em><sup>c</sup><sub>0</sub> be any orthogonal complement of <em>X</em><sub>0</sub>. Then</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e004.png" alt="equation" /></p> <p>is continuous and well defined for <em>X</em> ∈ <em>N</em> .</p> <p><strong>Proof: </strong>By Lemma 1 all <em>X</em> ∈ <em>N</em> have full column rank. Thus F(<em>X</em>) is well defined. It follows from the continuity of matrix multiplication and matrix inversion that <em>F</em>(<em>X</em>) is continuous for all <em>X</em> ∈ <em>N</em>.</p> <p><strong>Theorem 1:</strong> Let <em>X</em><sub>0</sub> be a <em>p</em>×<em>q</em> matrix with full column rank <em>q</em><<em>p</em>. Let <em>X</em><sup>c</sup><sub>0</sub> be any orthogonal complement of <em>X</em><sub>0</sub>. Then there is a neighbourhood N of <em>X</em><sub>0</sub> such that <em>X</em> has full column rank for all <em>X</em> ∈ <em>N</em> and</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e005.png" alt="equation" /></p> <p>is an orthogonal complement of <em>X</em>.</p> <p><strong>Proof:</strong> Note that <em>F</em>(<em>X</em>) is <em>p</em>×(<em>p</em>-<em>q</em>) and <em>F</em>(<em>X</em>) is orthogonal to <em>X</em>. It is sufficient to prove <em>F</em>(<em>X</em>) has full column rank. Let ɡ(<em>X</em> ) =det(<em>F</em>(<em>X</em> )′<em>F</em>(<em>X</em>) ) . Then ɡ( <em>X</em> ) is a continuous and it follows from this that there is <em>δ</em>>0 such that</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e006.png" alt="equation" /></p> <p>It follows from the last inequality that ɡ( <em>X</em> ) ≠ 0 . Hence det(<em>F</em>(<em>X</em> )′<em>F</em>(<em>X</em>) )≠ 0 and hence <em>F</em>(<em>X</em>) has full column rank.</p> <p>This proof is much simpler than that given by Jennrich and Satorra using the implicit function theorem [<a href="#1" title="1">1</a>]. In their paper Browne and Shapiro prove Browne's Proposition 4 without using their <em>F</em>(<em>X</em>) formula. One could, however, use Theorem 1 because it asserts the existence of a locally continuous orthogonal complement function and this is all that is needed to fix Browne's proof of Proposition 4. There are now three proofs of Browne's Proposition 4, the one given by Jennrich and Satorra [<a href="#1" title="1">1</a>], the one given by Browne and Shapiro [<a href="#3" title="3">3</a>], and the one using Theorem 1.</p> <p>A problem with using Theorem 1 is that for an <em>X</em> of interest one has no way of knowing if <em>X</em> is in <em>N</em> since the only thing we know about <em>δ</em> is that it is greater than zero. If necessary for an <em>X</em> of interest one can always compute an orthogonal complement of <em>X</em> using Theorem 1 of Jennrich and Satorra, but it will not be a continuous function of <em>X </em>[<a href="#1" title="1">1</a>].</p> <p>See the following example which shows numerically that Theorem 1 of Jennrich and Satorra produces orthogonal complements as does Theorem 1 of this document and also suggests an interesting alternative method using <em>F</em>(<em>X</em>) [<a href="#1" title="1">1</a>].</p> <p><strong>Theory:</strong></p> <p><strong>Example:</strong></p> <p>Let</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e007.png" alt="equation" /></p> <p>Then <em>X</em> is a <em>p</em> by <em>q</em> matrix with <em>p</em> = 4 and <em>q</em> = 2. Theorem 1 of Jennrich and Satorra [<a href="#1" title="1">1</a>] says that if </p> <p><em>X</em>= <em>QR</em></p> <p>is the long form of a QR factorization of X, then the last p-q columns of Q are an orthogonal complement of X. Computing a QR factorization of X gives</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e008.png" alt="equation" /></p> <p>The last <em>p</em>-<em>q</em> columns of this are:</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e009.png" alt="equation" /></p> <p>and this is an orthogonal complement <em>X</em>. This demonstrates numerically that Theorem 1 of Jennrich and Satorra [<a href="#1" title="1">1</a>] produces orthogonal complements.</p> <p>Note that when <em>X</em>=<em>X</em><sub>0</sub>, <em>X</em> satisfies the assumptions of Theorem 1 and the computed value of</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e010.png" alt="equation" /></p> <p>which is an orthogonal complement of <em>X</em>. Thus in this case at least <em>F</em>(<em>X</em>) is an orthogonal complement of <em>X</em> as asserted by Theorem 1.</p> <p>Let</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e011.png" alt="equation" /></p> <p>be a random matrix whose components are independent standard normal variables. Note that it has full column rank. The value of the Browne-Shapiro.</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e012.png" alt="equation" /></p> <p>This has full column rank and hence is an orthogonal complement of <em>X</em> even though <em>X</em> may not satisfy the basic assumption of Theorem 1 above because <em>X</em> is simply a random matrix.</p> <p>When this is repeated 1000 times in every case <em>F</em>(<em>X</em>) is an orthogonal complement of <em>X</em>. This suggests that <em>F</em>(<em>X</em>) is an orthogonal complement of <em>X</em> with high probability.</p> <p>We show below that much more is true. We show <em>F</em>(<em>X</em>) fails to be an orthogonal complement of <em>X</em> only for <em>X</em> in subset of Labesgue measure zero in <em>R<sup>pxq</sup></em>.</p> <p>Note that when <em>X</em> = <em>X</em><sup>c</sup><sub>0</sub> , F(<em>X</em>)=0 and hence <em>F</em>(<em>X</em>) is not an orthogonal complement of X for all X with full column rank.</p> <h2>A necessary and sufficient condition for F(X) to be an orthogonal complement of X.</h2> <p>To simplify the development that follows let us begin by looking at the simplest possible case when <em>p</em> = 2 and <em>q</em> = 1. Assume <em>X</em> is not in the column space of <em>X</em><sup>c</sup><sub>0</sub> . Then in <strong>Figure 1</strong> clearly, <em>F</em>(<em>X</em>) is a non-zero vector and has full column rank. Thus <em>F</em>(<em>X</em>) is an orthogonal complement of <em>X</em> for all <em>X</em> not in the column space <em>X</em><sup>c</sup><sub>0</sub> . If<em> X</em> is in the column space of <em>X</em><sup>c</sup><sub>0</sub> , then <em>F</em>(<em>X</em>)=0 and hence <em>F</em>(<em>X</em>) is not an orthogonal complement of <em>X</em>. Note this happens only on a set that has Labesgue measure zero in <em>R</em><sup>2</sup>. We will show that this happens for arbitrary <em>p</em> by <em>q</em><<em>p</em> matrices <em>X</em>.</p> <div class="well well-sm"> <div class="row"> <div class="col-xs-12 col-md-2"><a onclick="openimage('https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-orthogonal-2-1-47-g001.png','','scrollbars=yes,resizable=yes,width=500,height=330')" class="thumbnail"><img src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-orthogonal-2-1-47-g001.png" class="img-responsive" alt="applied-mathematics-statistical-applications-orthogonal" title="applied-mathematics-statistical-applications-orthogonal" /></a></div> <div class="col-xs-12 col-md-10"> <p><em><strong>Figure 1.</strong> Showing an orthogonal complement of X.</em></p> </div> </div> </div> <p>We begin with the following theorem,</p> <p><strong>Theorem 2:</strong> Let <em>F</em>(<em>X</em>) be the Browne-Shapiro formula. That [<em>X</em>,<em>X</em><sup>c</sup><sub>0</sub>] is invertible is a necessary and sufficient condition for <em>F</em>(<em>X</em>) to be an orthogonal complement of <em>X</em>.</p> <p><strong>Proof:</strong> Assume [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>] has full column rank. Assume <em>F</em>(<em>X</em>) does not have full column rank. Then</p> <p><em>F</em>(<em>X</em>)a=0</p> <p>for some vector a <em>a</em> ≠ 0. Thus</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e013.png" alt="equation" /></p> <p>It follows that<em> X</em><sup>c</sup><sub>0</sub>a is in the column space of <em>X</em>. Thus</p> <p><em>X</em><sup>c</sup><sub>0</sub>a = Xb</p> <p>for some vector b. Thus there is a vector in the column space of <em>X</em> that is in the column space of <em>X</em><sup>c</sup><sub>0</sub> . It follows that [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>] does not have full column rank. This contradiction proves <em>F</em>(<em>X</em>) has full column rank when [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>] has full column rank. This is a sufficient condition for <em>F</em>(<em>X</em>) to be an orthogonal complement of <em>X</em>.</p> <p>If [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>] does not have full column rank, then Xa = <em>X</em><sup>c</sup><sub>0</sub>b for two vectors a and b not both zero. If <em>a</em> = 0, then <em>b</em> = 0. This implies <em>a</em> ≠ 0 which in turn implies <em>b</em> ≠ 0. Now</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e014.png" alt="equation" /></p> <p>which implies that <em>F</em>(<em>X</em>) does not have full column rank. Thus <em>F</em>(<em>X</em>) is not an orthogonal complement of <em>X</em> when [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>] does not have full column rank. Thus <em>F</em>(<em>X</em>) is an orthogonal complement of <em>X</em> if and only if [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>] has full column rank.</p> <h2>How often does F(X) fail to be an orthogonal complement of X?</h2> <p><strong>Lemma 2:</strong> A polynomial in <em>n</em> variables is either identically 0 or its roots have Labesgue measure zero in <em>R<sup>n</sup></em>.</p> <p><strong>Proof:</strong> The proof is by mathematical induction on <em>n</em>. For<em> n</em>=1, <em>P</em>(<em>x</em>) of degree <em>d</em> can have at most d roots, which gives us the base case. Now, assume that the theorem holds for all polynomials in <em>n</em>-1 variables. Let <img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e015.png" alt="equation" /> and write</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e016.png" alt="equation" /></p> <p>Let <em>I</em>[<em>A</em>] be the indicator function of a set <em>A</em>. By Fubini's theorem,</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e017.png" alt="equation" /></p> <p>The first two integral equalities follow from Fubini's Theorem. The last integral equality follows from the fact that given <img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e018.png" width="87" height="26" /> is a polynomial in <em>n</em>-1 variables and by the induction hypothesis this is identically zero for all <img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e019.png" alt="equation" /> or its zeros have Labesgue measure zero in R<sup><em>n</em>-1</sup>[<a href="#4" title="4">4</a>].</p> <p><strong>Comment:</strong> Lemma 2 is a simple very interesting result which apparently is not in the literature. It is not in any journals covered in Jstor. It is not in any of the five analysis books we own and according to Caron and Traynor it is not in any measure theory book.</p> <p><strong>Theorem 3:</strong> The <em>X</em> for which the Browne - Shapiro function <em>F(X)</em> is not an orthogonal complement of <em>X</em> have Labesgue measure zero in <em>R<sup>p×q</sup></em><sup></sup>.</p> <p><strong>Proof: </strong>F(X) is not an orthogonal complement of X if and only if [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>] is singular. This happens if and only if det( [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>][<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>])=0 . As a function of X this is not identically zero because it is not zero when X=X<sub>0</sub>. Moreover det det( [<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>][<em>X</em>,<em> X</em><sup>c</sup><sub>0</sub>]) is a polynomial in <em>p×q </em>variables <em>x</em>=<em>vec</em>(<em>X</em>) which is not identically zero. It follows from Lemma 2 that its zeros have Lebesgue measure zero in Rpq. Or equivalently the <em>X</em> such that <em>F</em>(<em>X</em>) is not an orthogonal complement of <em>X</em> have Labesgue measure zero in <em>R<sup>p×q</sup></em>.</p> <p><strong>Remark: </strong>Theorem 3 implies that <em>F(X)</em> almost never fails to be an orthogonal complement of <em>X</em>. If <em>X</em> is a sample from a density, then <em>F(X)</em> is an orthogonal complement of <em>X</em> with probability one.</p> <h2>Conclusion</h2> <p>Browne and Sharpiro [<a href="#3" title="3">3</a>] give a formula for computing orthogonal complements. Let <em>X</em><sub>0</sub> be any <em>p</em>×<em>q</em> matrix with full column rank <em>q</em><<em>p</em> and <em>X</em><sup>c</sup><sub>0</sub> be any orthogonal complement of <em>X</em><sub>0</sub>. Browne and Sharpiro function is</p> <p><img class="equation" src="https://www.alliedacademies.org/articles-images/applied-mathematics-statistical-applications-2-1-47-e020.png" alt="equation" /></p> <p>They state that if <em>N</em> be any neighbourhood of <em>X</em><sub>0</sub> such that <em>X</em> has full column rank for all <em>X</em> ϵ <em>N</em> then <em>F</em>(<em>X</em>) is an orthogonal complement of <em>X</em> for all <em>X</em> in <em>N</em>. They, however, fail to show this. In particular they fail to show <em>F</em>(<em>X</em>) has full column rank. As a consequence their claim is only a conjecture. The main point of our paper is to prove their conjecture and much more. We prove that on all but a set of <em>p</em>×<em>q</em> matrices <em>X</em> with Labesque measure zero, <em>F</em>(<em>X</em>) is an orthogonal complement of <em>X</em>. In particular if <em>X</em> is a sample from a density,<em> F</em>(<em>X</em>) is an orthogonal complement of <em>X</em> with probability one.</p> <p>Theorem 1 is a proof for the Browne-Sharpiro conjecture. We also prove additional results that should enhance considerably the practical relevance of the Browne-Shapiro function for an orthogonal complement. Theorem 2 gives a necessary and sufficient condition for <em>F</em>(<em>X</em>) to be an orthogonal complement of <em>X</em>. Lemma 2 is a not well known, but very useful result about the roots of an arbitrary polynomial in n variables. If the polynomial is not identically zero, its roots have Labesque measure zero. Theorem 3 is our main result.</p> <h2>Acknowledgements</h2> <p>The authors would like to thank Ellen Jennrich for proof reading our paper and helping with its submission. The research of the second author is supported by grant EC02014-59885-P from the Spanish Ministry of Science and Innovation.</p> <h2 class="post-title">References</h2> <ol><li id='Reference_Titile_Link' value='1'><a name=1 id='1'></a>Jennrich RI, Satorra A. Continuous orthogonal complement functions and distribution free goodness of fit tests for covariance structure analysis. Psychometrika. 2013;78: 445-552.</a></li><li id='Reference_Titile_Link' value='2'><a name=2 id='2'></a>Browne MW. Asymptotically distribution-free methods for the analysis of covariance structures. Psychometrika. 1984;37: 62-83.</a></li><li id='Reference_Titile_Link' value='3'><a name=3 id='3'></a>Browne MW, Shapiro A. A comment on the asymptotics of a distribution free goodness of fit test statistic. Psychometrika. 2013;78: 196-9.</a></li><li id='Reference_Titile_Link' value='4'><a name=4 id='4'></a>Caron R, Traynor T. The zero set of a polynomial. 2005.</a></li></ol> </article> </div> </main>  <aside class="col-12 col-sm-3 order-sm-1"> <a href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications-online-visitors-readers-25.html" class="btn btn-danger btn-block mb-3 font-size-4"><i class="fas fa-book-reader"></i> 25+ Million Readerbase</a> <a class="btn btn-danger btn-block mb-3 font-size-4" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/nominations.php" title="Awards & Nominations">Awards & Nominations</a>  <div class="text-center mb-3"> <a class="btn btn-outline-dark btn-sm" href="https://www.alliedacademies.org/articles/more-on-the-orthogonal-complement-functions.pdf" title="View PDF"><i class="ti-eye"></i> View PDF</a> <a class="btn btn-outline-dark btn-sm" href="https://www.alliedacademies.org/download.php?download=articles/more-on-the-orthogonal-complement-functions.pdf" title="Download PDF"><i class="ti-cloud-down"></i> Download PDF</a> </div> <div class="card shadow-sm sidebar mb-0"> <div class="list-group list-group-flush qr_code_image"> <img title="QR" src="https://chart.googleapis.com/chart?chs=185x185&cht=qr&chl=https%3A%2F%2Fwww.alliedacademies.org%2Farticles%2Fmore-on-the-orthogonal-complement-functions-11697.html&chld=M|0&icqrf=00b1e4" alt="More on the orthogonal complement functions" />  <nav class="nav nav-pills social-icons-footer sidebar_social_icons a-pl-0">        <a style="width:30px;" title="Share This Article" target="_blank" class="nav-link" href="https://www.facebook.com/sharer.php?s=100&p[title]=More on the orthogonal complement functions&p[url]=https%3A%2F%2Fwww.alliedacademies.org%2Farticles%2Fmore-on-the-orthogonal-complement-functions-11697.html"><img src="https://www.alliedacademies.org/assets/socials/facebook.png" alt="More on the orthogonal complement functions" /></a>      <a style="width:30px;" target="_blank" href="https://web.whatsapp.com/send?text=https%3A%2F%2Fwww.alliedacademies.org%2Farticles%2Fmore-on-the-orthogonal-complement-functions-11697.html" title="Share This Article" class="nav-link"><img src="https://www.alliedacademies.org/assets/socials/whatsapp.png" alt="More on the orthogonal complement functions" /></a>      <a style="width:30px;" target="_blank" href="https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.alliedacademies.org%2Farticles%2Fmore-on-the-orthogonal-complement-functions-11697.html" title="Share This Article" class="nav-link"><img src="https://www.alliedacademies.org/assets/socials/linkedin.png" alt="More on the orthogonal complement functions" /></a>      <a style="width:30px;" title="Share This Article" target="_blank" href="https://twitter.com/share?text=More on the orthogonal complement functions&url=https%3A%2F%2Fwww.alliedacademies.org%2Farticles%2Fmore-on-the-orthogonal-complement-functions-11697.html" class="nav-link"><img src="https://www.alliedacademies.org/assets/socials/twitter.png" alt="More on the orthogonal complement functions" /></a> </nav>  </div> </div> <div class="cu_navgation"> <nav class="sidebar-nav mb-3 p-0"> <a class="navbar-brand d-block d-sm-none" href="#">Journal Menu</a> <div> <div class="navbar-nav flex-column w-100"> <a class="nav-item nav-link active" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/" title="Journal Home">Journal Home</a> <a class="nav-item nav-link" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/editors.php" title="Editorial Board"><i class="ti-id-badge"></i> Editorial Board</a> <a class="nav-item nav-link" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/guidelines.php" title="Author Guidelines"><i class="ti-pencil-alt"></i> Author Guidelines</a> <a class="nav-item nav-link" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/submit-manuscript.php" title="Submit Manuscript"><i class="ti-hand-point-up"></i> Submit Manuscript</a> <a class="nav-item nav-link" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/inpress.php" title="In Press"><i class="ti-files"></i>Articles in process</a>  <a class="nav-item nav-link" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications/volume-selector.php" title="Volume Selector"><i class="ti-layers-alt"></i> Volume Selector</a> <a class="nav-item nav-link" href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications-advertising-80.html" title="Advertise"><i class="ti-files"></i> Advertise</a> <a class="nav-item nav-link" href="https://www.alliedacademies.org/covid-19-peer-reviewed-journals-articles-special-issues.php" title="Advertise"><i class="ti-layers-alt"></i> COVID-19 Special Issues</a> </div> </div> </nav> </div>       <div class="card mb-3 radius-0 shadow-sm"> <div class="card-body font-size-18 radius-0 py-2 fweight-500"> Latest Volumes </div> <div class="card-header"> <select class="form-control" id="year" name="year"> <option value="2019" selected = selected > 2019</option> <option value="2018" > 2018</option> </select> </div> <div class="nav flex-column font-size-16 py-2" id="vol_disp"> </div> </div>  <div class="card mb-3"> <a href="https://www.alliedacademies.org/journal-applied-mathematics-statistical-applications-advertising-80.html" title="Click here"><img src="https://www.alliedacademies.org/assets/images/tower-banner.jpg" alt="Flyer image" class="img-fluid p_rel" /> <span class="p_abo cu_roundchip"> <span> <h5><span>25+</span> Million Website Visitors</h5> </span> </span> </a> </div>      <div class="card shadow-sm sidebar mb-3"> <h6 class="card-header pr-0">Open Access Journals</h6> <div class="list-group list-group-flush"> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/chemistry-journals.php" title="Chemistry" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Chemistry</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/clinical-journals.php" title="Clinical Sciences" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Clinical Sciences</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/food-nutrition-journals.php" title="Food & Nutrition" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Food & Nutrition</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/general-science-journals.php" title="General Science" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>General Science</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/genetics-molecular-biology-journals.php" title="Genetics & Molecular Biology" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Genetics & Molecular Biology</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/immunology-microbiology-journals.php" title="Immunology & Microbiology" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Immunology & Microbiology</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/medical-journals.php" title="Medical Sciences" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Medical Sciences</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/neurology-psychology-journals.php" title="Neurology & Psychology" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Neurology & Psychology</a> </div> <div class="list-group-item p-0 pt-2 pl-2"> <a class="pr-2 deep-orange-400-before cu_bu" href="https://www.alliedacademies.org/nursing-health-care-journals.php" title="Nursing & Health Care" target="_blank"><span class="text-danger mr-1"><i class="far fa-hand-point-right"></i></span>Nursing & Health Care</a> </div> </div> </div>  </aside>  <script type="text/javascript" src="https://ajax.googleapis.com/ajax/libs/jquery/1.4.2/jquery.min.js"></script> <script type="text/javascript"> $(document).ready(function() { $("#year").val('2019'); yearchange('2019'); $("#year").change(function() { var year=$("#year").val(); yearchange(year); }); function yearchange(year){ //var year=$("#year").val(); var journal="aaams"; var dataString = {'year':year,'journal':journal}; $.ajax ({ type: "POST", url: "https://www.alliedacademies.org/vol_disp.php", data: dataString, cache: false, success: function(html) { $("#vol_disp").html(html); } }); } }); </script> </div> </div> </section> <footer class="bg-grey-800 font-size-14">  <a href="#0" class="cd-top js-cd-top"> <svg viewBox="0 0 80 60" preserveAspectRatio="none" class="back-to-top-svg"><use xlink:href="#tabshape"></use></svg> <svg viewBox="0 0 80 60" preserveAspectRatio="none" class="back-to-top-svg"><use xlink:href="#tabshape"></use></svg> <span><i class="fas fa-angle-up"></i></span> </a> <div class="container"> <div class="row align-items-start border-bottom-1 brd-grey-700"> <div class="col-xs-12 col-sm-3"> <h5 class="mt-2">Quick Links</h5> <nav class="nav flex-column"> <a class="nav-link grey-100 pl-0 py-1" href="https://www.alliedacademies.org/">Home</a> <a class="nav-link grey-100 pl-0 py-1" href="https://www.alliedacademies.org/journals.php">Open Acces Journals</a> <a class="nav-link grey-100 pl-0 py-1" href="https://www.alliedacademies.org/journal-submission-instructions.php">Submission Instructions</a> <a class="nav-link grey-100 pl-0 py-1" href="https://www.alliedacademies.org/publication-guidelines.php">Publication Guidelines</a> <a class="nav-link grey-100 pl-0 py-1" href="https://www.alliedacademies.org/contact.php">Contact</a> </nav> </div> <div class="col-xs-12 col-sm-3"> <h5 class="mt-2">Reach Us</h5> <address> Allied Academies<br> 40 Bloomsbury Way<br> Lower Ground Floor<br> London, United Kingdom<br> WC1A 2SE </address> </div> <div class="col-xs-12 col-sm-2"> <h5 class="mt-2">Contact Us</h5> <p> Call: 44 2033180199<br> E-Mail: <a href="mailto:info@alliedacademies.org" class="grey-100">info@alliedacademies.org</a> </p> </div> <div class="col-xs-12 col-sm-3"> <h5 class="mt-2 text-center">Follow Us</h5> <ul class="list-inline mb-0 text-center"> <li class="animate-icon list-inline-item mt-0"> <a href="https://www.facebook.com/alliedacdemies/" class="animate-icon-wrap animate-icon-sm bg-facebook-hover rounded-circle" target="_blank" title="Facebook"> <i class="animate-icon-item fab fa-facebook-f white"></i> <i class="animate-icon-item fab fa-facebook-f white"></i> </a> </li> <li class="animate-icon list-inline-item mt-0"> <a href="https://twitter.com/AlliedAcadamies" class="animate-icon-wrap animate-icon-sm white bg-twitter-hover rounded-circle" target="_blank" title="Twitter"> <i class="animate-icon-item fab fa-twitter"></i> <i class="animate-icon-item fab fa-twitter"></i> </a> </li> <li class="animate-icon list-inline-item mt-0"> <a href="https://www.linkedin.com/company/allied-academies" class="animate-icon-wrap animate-icon-sm white bg-linkedin-hover rounded-circle" target="_blank" title="Linkedin"> <i class="animate-icon-item fab fa-linkedin-in"></i> <i class="animate-icon-item fab fa-linkedin-in"></i> </a> </li> <li class="animate-icon list-inline-item mt-0"> <a href="https://www.instagram.com/allied_academies9/" class="animate-icon-wrap animate-icon-sm white bg-instagram-hover rounded-circle" target="_blank" title="Instagram"> <i class="animate-icon-item fab fa-instagram"></i> <i class="animate-icon-item fab fa-instagram"></i> </a> </li> </ul> </div> </div> <div class="row"> <div class="col"> <p class="my-3">© Allied Academies 2025 | Creative Commons License Open Access Journals by Allied Academies is licensed under a Creative Commons Attritbution 4.0 International License</p> </div> </div> </div> </footer> <div class="app-view"> Get the App <p><a href="https://play.google.com/store/apps/details?id=com.journals.alliedacademies" target="_blank"> <i class="fab fa-android"></i></a></p> </div>    <script src="https://code.jquery.com/jquery-3.3.1.min.js"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/jqueryui/1.12.1/jquery-ui.js"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.14.3/umd/popper.min.js"></script> <script src="https://stackpath.bootstrapcdn.com/bootstrap/4.1.1/js/bootstrap.min.js"></script>  <script src="/assets/js/scroll-to-top.js"></script> <script> $(function () { $('[data-toggle="tooltip"]').tooltip() }) </script>  <script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script> <script language="javascript"> $( "#search_key" ).autocomplete({ source: function(request, response) { $.ajax({ url: "https://www.alliedacademies.org/author_names.php", type: "POST", dataType: "json", data: { keyword: request.term, }, success: function( data ) { response( data ); } }); }, select: function (event, ui) { $("#search_key").val(ui.item.value); $("#keyword").val(ui.item.id); location.href = 'https://www.alliedacademies.org/search-results.php?keyword='+encodeURI(ui.item.value); }, change: function (event, ui) { var titletext = $('#search_key').val(); if(titletext == ""){ $('#keyword').val(''); } }, minLength: 2, max:50, highlightClass: "bold-text", scroll:true }); </script> <script type="text/javascript" src="https://www.alliedacademies.org/js/jquery.back-to-top.js"></script> <script type="text/javascript" src="https://www.alliedacademies.org/js/dzsparallaxer.js"></script> <script src="//cdnjs.cloudflare.com/ajax/libs/metisMenu/2.7.7/metisMenu.min.js"></script>  <script> $("#menu").metisMenu({ triggerElement: '.nav-link', parentTrigger: '.nav-item', subMenu: '.nav.flex-column' }); </script>  </body> </html>