B Z;m@spdZddlmZddlZddlmZddlmZm Z ddl m Z ddZ dd d Zd dZdddZdddZdddZd dddgZedkrleddZeeZd Zeddddddddd d!d"d#g Zeddgd$d%gd&d'gd&d'gd&d'gd&d'gd&d'gd&d'gd&d'gd&d'gd$d%gddgg Zejed(efZeeeed d)eeeed d)ed*d+eeeDeed,d d d)Z eed(d d d)Z!ee ee!ee!e e e eed,ddfeed d-d)e eed(e d(d,dfeed d.d)e ede d,dfeeed d)ed/ejed(efZ"eee"d,d d.d)eee"ed d)eee"ed d)e eed,dddfee"d d-d)e eed(e d(d,ddfee"d d.d)e ede d,ddfee"ed d)e eed,dee"d,d d-d)e eed(e d(d,ee"d,d d.d)e ede d,ee"ed d)dd0lm#Z#ee#j$j%e"ed,d,d,gddd1edd2d3d4dd5d6d7d8d9d:d;dd?d@dAdBdCdDdEdFdGdHdIdJdKdLdMdNdOdPdQdRdSdTdUdVdWdXdYdZd[d\d]d^d_d`dadbdcdddedfdgdhdidjdkdldmdndodpdqdrdsdtdudvdwdxdydzd{d|d}d~dddddddddddddddddddddgdZ&e e&eeddd ed3d4dd5d6d7d8d9d:d;dd?d@dAdBdCdDdEdFdGdHdIdJdKdLdMdNdOdPdQdRdSdTdUdVdWdXdYdZd[d\d]d^d_d`dadbdcdddedfdgdhdidjdkdldmdndodpdqdrdsdtdudvdwdxdydzd{d|d}d~ddddddddddddddddddddddddddddddgkZ'e e'eeddd-edddddddd d!dddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddgdZ(e e(eeddd.dS(ausing scipy signal and numpy correlate to calculate some time series statistics original developer notes see also scikits.timeseries (movstat is partially inspired by it) added 2009-08-29 timeseries moving stats are in c, autocorrelation similar to here I thought I saw moving stats somewhere in python, maybe not) TODO moving statistics - filters don't handle boundary conditions nicely (correctly ?) e.g. minimum order filter uses 0 for out of bounds value -> append and prepend with last resp. first value - enhance for nd arrays, with axis = 0 Note: Equivalence for 1D signals >>> np.all(signal.correlate(x,[1,1,1],'valid')==np.correlate(x,[1,1,1])) True >>> np.all(ndimage.filters.correlate(x,[1,1,1], origin = -1)[:-3+1]==np.correlate(x,[1,1,1])) True # multidimensional, but, it looks like it uses common filter across time series, no VAR ndimage.filters.correlate(np.vstack([x,x]),np.array([[1,1,1],[0,0,0]]), origin = 1) ndimage.filters.correlate(x,[1,1,1],origin = 1)) ndimage.filters.correlate(np.vstack([x,x]),np.array([[0.5,0.5,0.5],[0.5,0.5,0.5]]), origin = 1) >>> np.all(ndimage.filters.correlate(np.vstack([x,x]),np.array([[1,1,1],[0,0,0]]), origin = 1)[0]==ndimage.filters.correlate(x,[1,1,1],origin = 1)) True >>> np.all(ndimage.filters.correlate(np.vstack([x,x]),np.array([[0.5,0.5,0.5],[0.5,0.5,0.5]]), origin = 1)[0]==ndimage.filters.correlate(x,[1,1,1],origin = 1)) update 2009-09-06: cosmetic changes, rearrangements )print_functionN)signal)assert_array_equalassert_array_almost_equalcCsP|}t|dkr$|t|df}tjt||d|t||dfS)Nr)npndimshapeZr_ones)xkZkaddr>lib/python3.7/site-packages/statsmodels/sandbox/tsa/movstat.py expandarr6srmedlaggedcCs|dkr|d}n*|dkr d}n|dkr8| dd}ntt|dkrP|}n:|dkrf|dd}n$|d krtd}n|d kr|d}ntt||}t|t|||||| S) almoving order statistics Parameters ---------- x : array time series data order : float or 'med', 'min', 'max' which order statistic to calculate windsize : int window size lag : 'lagged', 'centered', or 'leading' location of window relative to current position Returns ------- filtered array rrcenteredrleadingrTrminmax) ValueErrorr ZisfiniterrZ order_filterr )r orderwindsizelagleadordxextrrrmovorder=s$   r cCs6ddlm}tdd}t|dd}t||tddd}t|dd}t||tt|dd d dd|ddtdd tjd }t|d}t|dd}| | ||d ||d | dt|dd d }| | ||d ||d | dt|ddd }| | ||d ||d | ddS)zgraphical test for movorderrNr r)rrrr)rrrz.-zmoving max laggedzmoving max centeredrzmoving max leading) Zmatplotlib.pylabZpylabr aranger rZlinspaceZpiZsinZfigureZplottitle)Zpltr ZxoZttrrrcheck_movorderks,      $   r%cCst|d||dS)amoving window mean Parameters ---------- x : array time series data windsize : int window size lag : 'lagged', 'centered', or 'leading' location of window relative to current position Returns ------- mk : array moving mean, with same shape as x Notes ----- for leading and lagging the data array x is extended by the closest value of the array r) windowsizer) movmoment)r r&rrrrmovmeansr(cCs,t|d||d}t|d||d}|||S)aCmoving window variance Parameters ---------- x : array time series data windsize : int window size lag : 'lagged', 'centered', or 'leading' location of window relative to current position Returns ------- mk : array moving variance, with same shape as x r)r&rr)r')r r&rm1m2rrrmovvarsr+c CsT|}|dkr0d}t|dpdd|dp*d}n|dkrp| d}t|d|dpVd|d |dpjd}nL|dkr| d}td|dd|pdd|d| dpd}ntt|t|}t||d}t||jdkrt|||d |St|j t|dddfj t |||dddfd |ddfSdS) anon-central moment Parameters ---------- x : array time series data windsize : int window size lag : 'lagged', 'centered', or 'leading' location of window relative to current position Returns ------- mk : array k-th moving non-central moment, with same shape as x Notes ----- If data x is 2d, then moving moment is calculated for each column. rrrNrrrZfull) slicerr r floatrprintr Z correlater r) r rr&rrrZslZavgkernrrrrr's&  . 6  r'__main__z! checkin moving mean and variancer!ggUUUUUU?g?g@g@g@g@g@g@g @gUUUUUU!@ g#q?g| q?gvWUU?gUUU@r)r&rcCs"g|]}tt|t|qSr)r varr ws).0irrr ?sr6rrrz- checking moving moment for 2d (columns only))ndimage)Zaxisg?g333333?g333333?g?g@gffffff@g @g@g@g@g@g!@g#@g%@g'@g)@g+@g-@g/@g0@g1@g2@g3@g4@g5@g6@g7@g8@g9@g:@g;@g<@g=@g>@g?@g@@@g@@g@A@gA@g@B@gB@g@C@gC@g@D@gD@g@E@gE@g@F@gF@g@G@gG@g@H@gH@g@I@gI@g@J@gJ@g@K@gK@g@L@gL@g@M@gM@g@N@gN@g@O@gO@g P@g`P@gP@gP@g Q@g`Q@gQ@gQ@g R@g`R@gR@gR@g S@g`S@gS@gS@g T@g`T@gT@gT@g U@g`U@gU@gU@g V@g`V@gV@gV@g W@g`W@gW@dgW@g X@g9X@g`X@gX@gX@g̬X@gX@gX@g@g?@g@@g@@gA@gA@gB@gB@gC@gC@gD@gD@gE@gE@gF@gF@gG@gG@gH@gH@gI@gI@gJ@gJ@gK@gK@gL@gL@gM@gM@gN@gN@gO@gO@gP@g@P@gP@gP@gQ@g@Q@gQ@gQ@gR@g@R@gR@gR@gS@g@S@gS@gS@gT@g@T@gT@gT@gU@g@U@gU@gU@gV@g@V@gV@gV@gW@g@W@gW@g⣋.W@g٧뢋W@gEX@gb\tEX@gw.hX@ )rrr)rr)rr)rr))__doc__Z __future__rZnumpyr ZscipyrZ numpy.testingrrZstatsmodels.apiZapiZsmrr r%r(r+r'__all____name__r/Znobsr#r r3ZarrayZaveZvaZc_Zave2dranger)r*Zx2dr7filtersZ correlate1dZxgZxdZxcrrrr+s    .5   S      " &