<record>
    <language>eng</language>
    <publisher>TQMP</publisher>
    <journalTitle>The Quantitative Methods for Psychology</journalTitle>
    <eissn>1913-4126</eissn>
    <publicationDate>2016-09-15</publicationDate>
    <volume>12</volume>
    <issue>2</issue>
    <startPage>131</startPage>
    <endPage>137</endPage>
	<doi>10.20982/tqmp.12.2.r131</doi>
    <documentType>article</documentType>
    <title language="eng">Le quotient de deux variances corrélées, sa distribution et son test</title>

    <authors>
      <author>
        <name>Laurencelle, Louis</name>
        <email>louis.laurencelle@gmail.com</email>
        <affiliationId>a</affiliationId>
      </author>
    </authors>

    <affiliationsList>
      <affiliationName affiliationId="1">Université du Québec à Trois-Rivières</affiliationName>
    </affiliationsList>

    <abstract language="eng">
       The joint sampling distribution of two correlated variances, i.e. variances stemming from a bivariate normal distribution or from two normal $\rho $-correlated distributions, is hardly known and used, by contrast with the distribution of $F$, the quotient of two independent, zero-correlated variances. The distribution of $F_\rho $, the quotient of two correlated variances, established by Bose (1935) and Finney (1938), is given along with its main characteristics, to which is added a handy $F_\rho $ to $F$ transformation. Finally, data based on Monte Carlo simulations document and compare the accuracy and power of two approximate tests of the difference between two correlated sample variances.  
    </abstract>

    <fullTextUrl format="pdf">https://www.tqmp.org/RegularArticles/vol12-2/p131/p131.pdf</fullTextUrl>

    <keywords language="eng">    
      <keyword>Correlated variances</keyword>
      <keyword>t-test on correlated variances</keyword>
    </keywords>
  </record>