Forskjell mellom versjoner av «CorrelationMatrix»
Fra mn/fys/epf
(Ny side: The covariance matrix is <br> <math>C_{ij} = <(x_i-\bar{x_i})(x_j-\bar{x_j})></math>.<br> Suppose <math>\delta_1,\ \delta_2,\ \delta_3</math> are three independent sources of normally-distr...) |
|||
Linje 1: | Linje 1: | ||
− | The covariance matrix is <br> | + | The covariance matrix is <br> <math>C_{ij} = <(x_i-\bar{x_i})(x_j-\bar{x_j})></math>.<br> Suppose <math>\delta_1,\ \delta_2,\ \delta_3</math> are three independent sources of normally-distributed unit fluctuations (with <span class="texhtml"> < δ<sub>''i''</sub> > = 0</span> and <span class="texhtml"> < δ<sub>''i''</sub> * δ<sub>''j''</sub> > = δ<sub>''i''''j'''</sub></span>'''and where <span class="texhtml">δ<sub>''i'''</sub></span>'''''j'' is the kronikker delta function (1 for <span class="texhtml">''i'' = ''j''</span> and zero for <math>i\neq j</math>). |
− | <math>C_{ij} = <(x_i-\bar{x_i})(x_j-\bar{x_j})></math>.<br> | ||
− | Suppose <math>\delta_1,\ \delta_2,\ \delta_3</math> are three independent sources of normally-distributed unit fluctuations (with < | ||
− | and < | ||
− | + | <br> Pseudo-measurements of <span class="texhtml">(''x''<sub>1</sub>,''x''<sub>2</sub>)</span> can be generated from the expressions <span class="texhtml">(''x''<sub>1</sub>,''x''<sub>2</sub>) = (''x''<sub>10</sub> + α * δ<sub>1</sub> + β * δ<sub>2</sub>,''x''<sub>20</sub> + γ * δ<sub>3</sub> + λ * δ<sub>2</sub>)</span>. | |
− | Pseudo-measurements of < | + | <blockquote>Expanding the covariance matrix one finds <math>C_{ij}=<x_i*x_j>-\bar{x_i}\bar{x_j}<math>.<br> |
− | + | ||
− | Expanding the covariance matrix one finds | + | Using the properties of <math><\delta_i*\delta_j></math> above and <math><x_i>=\bar{x_i}</math>, one finds <x_1*x_2> = x_{10}*x_{20}+\beta*\lambda</math> so that <span class="texhtml">''C''<sub>12</sub> = ''C''<sub>21</sub> = β * λ</span>. Similar substitution for <span class="texhtml">''C''<sub>''i''''j'''</sub></span>'''gives for the diagonal matrix elements <math>C_{11}=x_{10}^2+\alpha^2+\beta^2-x_{10}^2</math><br> and <br> <math>C_{22}=x_{20}^2+\gamma^2+\lambda^2-x_{10}^2</math>. so that C= ''' |
− | <math>C_{ij}=<x_i*x_j>-\bar{x_i}\bar{x_j}< | + | {| width="200" cellspacing="1" cellpadding="1" border="1" |
− | Using the properties of <math><\delta_i*\delta_j></math> above and <math><x_i>=\bar{x_i}</math>, one finds | + | |- |
− | <math>C_{11}=x_{10}^2+\alpha^2+\beta^2-x_{10}^2</math> and | + | | <span class="texhtml">α<sup>2</sup> + β<sup>2</sup></span> |
− | <math>C_{22}=x_{20}^2+\gamma^2+\lambda^2-x_{10}^2</math>. | + | | <span class="texhtml">β * λ</span> |
+ | |- | ||
+ | | <math>\beta*\lambda</math> | ||
+ | | <math>\gamma^2+\lambda^2</math> | ||
+ | |} | ||
+ | </blockquote> |
Revisjonen fra 25. mar. 2011 kl. 23:21
The covariance matrix is
.
Suppose are three independent sources of normally-distributed unit fluctuations (with < δi > = 0 and < δi * δj > = δi'jand where δij is the kronikker delta function (1 for i = j and zero for ).
Pseudo-measurements of (x1,x2) can be generated from the expressions (x1,x2) = (x10 + α * δ1 + β * δ2,x20 + γ * δ3 + λ * δ2).
Expanding the covariance matrix one finds C12 = C21 = β * λ. Similar substitution for Ci'jgives for the diagonal matrix elementsabove and , one finds <x_1*x_2> = x_{10}*x_{20}+\beta*\lambda</math> so that
and
. so that C=
α2 + β2 β * λ