U Kv f|ã@sddlZGdd„dƒZdS)éNc@seZdZdZdZdd„Zd$dd„Zdd „Zd d „Zd d „Z dd„Z dd„Z d%dd„Z d&dd„Z dd„Zdd„Zd'dd„Zd(d d!„Zd)d"d#„ZdS)*ÚPcazˆ A basic class for Principal Component Analysis (PCA). p is the number of dimensions, while N is the number of data points )ÚrÚgÚbÚcÚyÚmÚkcCs@|j}|tj|dd}|tj|dd}||_||_d|_dS)Nr©Zaxis)ÚAÚnpÚmeanÚstdÚMÚNÚ_eig)Úselfr rr©rú;/tmp/pip-unpacked-wheel-2v6byqio/statsmodels/sandbox/pca.pyZ__calc s z Pca.__calcNcCsÈt |¡j}|j\}}|||_|_||krBddlm}|dtƒ||_ |  ¡|_ |  ¡t  |jt|dt|jƒƒd¡d|…|_|dk r¤t|ƒ|kr¤tdƒ‚|dkr°dntdd„|Dƒƒ|_dS) z$ p X N matrix input r)Úwarnzp > n - intentional?éNznames must match data dimensioncSsg|] }t|ƒ‘qSr)Ústr)Ú.0ÚxrrrÚ (sz Pca.__init__..)r ÚarrayÚTÚshapeÚnÚpÚwarningsrÚRuntimeWarningr ÚcopyÚ_origAÚ _Pca__calcZtileÚ_colorsÚintÚlenÚ ValueErrorÚtupleÚnames)rÚdatar*r rrrrrrÚ__init__s     .z Pca.__init__cCst |jj¡S)z? returns the covariance matrix for the dataset )r Zcovrr©rrrrÚgetCovarianceMatrix+szPca.getCovarianceMatrixcCs^|jdkrXtj | ¡¡}t |d¡ddd…}|d||ddd…|ff}||_|jS)zQ returns a tuple of (eigenvalues,eigenvectors) for the data set. Nréÿÿÿÿr)rr ÚlinalgZeigr.Zargsort)rÚresZsortirrrÚgetEigensystem1s   zPca.getEigensystemcCs | ¡dS)Nr©r2r-rrrÚgetEigenvalues<szPca.getEigenvaluescCs | ¡dS)Nrr3r-rrrÚgetEigenvectors?szPca.getEigenvectorscCs| ¡}|t |¡S)z= "energies" are just normalized eigenvectors )r4r Úsum)rÚvrrrÚ getEnergiesBszPca.getEnergiesrrTc Csddlm}|jdd…|f|jdd…|f}}|r>| ¡| ||¡| ¡\}}| ¡\} } | ¡\} } | | | | } }t||j |j ƒD]B\}}}|j dd||||||d| |dd||dq|j dk r|  d|j |d¡| d|j |d¡dS) zü Generates a 2-dimensional plot of the data set and principle components using matplotlib. ix specifies which p-dimension to put on the x-axis of the plot and iy specifies which to put on the y-axis (0-indexed) rNgš™™™™™©?égà?)Z head_widthZfcZecú$z/\sigma$)Zmatplotlib.pyplotZpyplotrÚclfZscatterr2ZxlimZylimÚziprr%Zarrowr*ÚxlabelÚylabel)rÚixÚiyr;ZpltrrÚvalsZevsZxlZxuZylZyuZdxZdyÚvalZvecrrrrÚplot2dIs &    : z Pca.plot2déc CsÚddlmm}|r| ¡t d¡}| ¡| ¡}|j|||||||||dd|j |j dd…|f|j dd…|f|j dd…|fdd|j rÎ|j |j |d|j |d|j |ddn|  ¡dS) zõ Generates a 3-dimensional plot of the data set and principle components using mayavi. ix, iy, and iz specify which of the input p-dimensions to place on each of the x,y,z axes, respectively (0-indexed). rNéé)Z scale_factorg333333Ó?z/sigma)r=r>Zzlabel) Zenthought.mayavi.mlabZmayaviZmlabr;r Úzerosr5r4Zquiver3dZpoints3drr*Zaxes)rr?r@Zizr;rZz3r7rrrÚplot3dcs $<0z Pca.plot3dcCszt |¡r |t |jjd¡}|tj|jdd}|jjd}tjt |j¡|kdd}|j||_|  ¡|t |ƒS)a  clips out all data points that are more than a certain number of standard deviations from the mean. sigs can be either a single value or a length-p sequence that specifies the number of standard deviations along each of the p dimensions. rr r) r ZisscalarZonesrrrÚallÚabsr r$r6)rZsigsrrrrrÚsigclipws   z Pca.sigclipcCs|j ¡|_| ¡dS©N)r#r"r r$r-rrrÚreset‰s z Pca.resetcCsètdd„|||fDƒƒ}|dkr*tdƒ}n`|dkrthreshold) dimension array cSsg|] }|dk ‘qSrLr)rÚerrrr™szPca.project..rNrz&cannot specify more than one thresholdzShould be unreachableF)r"zshape for vals does not match) r6Úslicer(Zenergiesr ZcumsumÚ RuntimeErrorrrrrÚmatrixr5)rrAZenthreshZnPCsZcumenZnonnonesrÚprojrrrÚprojectŽs&     z Pca.projectc CsÀt |¡}|j\}}|jjd}||kr0tdƒ‚tj t | ¡¡j ¡}t  ||f¡}||dd…d|…f<||j }|r†t  |j ¡j Stj |jdd} tj |jdd} t  |j ¡| | j SdS)zP input is an n X q array, where q <= p output is p X n rzq > pNrr )r Z atleast_2drr r(r0ÚinvrQr5rrGrr rr) rr ÚnormedrÚqrZevinvZzsrRZmnsZsdsrrrÚ deproject±s    z Pca.deprojectcCs¢|dkr|j}n$|j}|jd|jjdkr4tdƒ‚| ¡}t |¡}|dd…|f|dd…|f<| |d¡}|j|}|jtj|j dd}|tj |jddS)z¯ pc can be a scalar or any sequence of pc indecies if vals is None, the source data is self.A, else whatever is in vals (which must be p x m) Nrz1vals do not have the correct number of componentsFrr ) r rrr(rSr Z zeros_likerWrrr )rZpcrAZpcsZzpcsZupcr ÚBrrrÚ subtractPCËs   zPca.subtractPC)N)rrT)rrrDT)NNNN)T)N)Ú__name__Ú __module__Ú __qualname__Ú__doc__r%r$r,r.r2r4r5r8rCrHrKrMrSrWrYrrrrrs      # r)Znumpyr rrrrrÚs