ó áp7]c@s dZddlmZddlmZmZmZmZmZddl Z ddl Z dZ iZ defd„ƒYZdefd „ƒYZd efd „ƒYZd efd „ƒYZed„Zdd„ZeddeƒZde_ddgd„Zd„ZdS(s© Provides the basic classes needed to specify statistical models. namespace : dictionary mapping from names to data, used to associate data to a formula or term iÿÿÿÿ(t OrderedDict(titerkeystlranget string_typest itervaluestrangeNtrestructuredtexttTermcBs€eZdZd„Zd d d„Zd„Zd„Zd„Ze eeeƒZ d„Z d„Z d„Z d „Zd „ZRS( sˆ This class is very simple: it is just a named term in a model formula. It is also callable: by default it namespace[self.name], where namespace defaults to formula.default_namespace. When called in an instance of formula, the namespace used is that formula's namespace. Inheritance of the namespace under +,*,- operations: ---------------------------------------------------- By default, the namespace is empty, which means it must be specified before evaluating the design matrix. When it is unambiguous, the namespaces of objects are derived from the context. Rules: ------ i) "X * I", "X + I", "X**i": these inherit X's namespace ii) "F.main_effect()": this inherits the Factor F's namespace iii) "A-B": this inherits A's namespace iv) if A.namespace == B.namespace, then A+B inherits this namespace v) if A.namespace == B.namespace, then A*B inherits this namespace Equality of namespaces: ----------------------- This is done by comparing the namespaces directly, if an exception is raised in the check of equality, they are assumed not to be equal. cs¡ytˆƒ‰Wntdƒ‚nXˆtˆƒkrTd|jtˆƒf}nd|jˆf}t|d|d‡fd†ƒ}ˆ|_|j|_|S(sd Raise the quantitative term's values to an integer power, i.e. polynomial. sexpecting a floats%s^%ds%s^%0.2ftfunct transformcstj|ˆƒS(N(tnptpower(tx(R (s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pytHt(tfloatt ValueErrortinttnamet QuantitativeR t namespace(tselfR Rtvalue((R s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt__pow__8s!  cCsj||_d|_|dkr*||_n ||_t|jtƒsTtdƒ‚n|rf||_ndS(Nsexpecting a string for termname(RtNonet_Term__namespacettermnamet isinstanceRRR(RRRR((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt__init__Ms     cCs-t|jtjƒr|jS|jp(tSdS(N(RRR tndarraytdefault_namespace(R((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt_get_namespace]scCs ||_dS(N(R(RR((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt_set_namespacebRcCs |`dS(N(R(R((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt_del_namespacecRcCs d|jS(s. '' % self.termname s (R(R((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt__str__fscCsGt|d|jƒ}||}t|j|jƒrC|j|_n|S(s0 Formula(self) + Formula(other) R(tFormulaRt_namespace_equal(Rtothertfothertf((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt__add__ls  cCs¤t|tƒr6|jdkr6t|d|jƒ}nF|jdkr]t|d|jƒ}nt|d|jƒ}||}t|j|jƒr |j|_n|S(s0 Formula(self) * Formula(other) t interceptR(RRRR#RR$(RR%R'((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt__mul__vs cCs-t|jtƒr|jgSt|jƒSdS(s† Return the names of the columns in design associated to the terms, i.e. len(self.names()) = self().shape[0]. N(RRRtlist(R((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pytnames†s cOs˜t|dƒs"|j|j}n |j}t|ƒr|t|ttfƒrjtj|ƒ}|j|_n|||Ž}nt j |ƒ}t j |ƒS(sé Return the columns associated to self in a design matrix. If the term has no 'func' attribute, it returns ``self.namespace[self.termname]`` else, it returns ``self.func(*args, **kw)`` R( thasattrRRRtcallableRRR#tcopyR tasarraytsqueeze(Rtargstkwtval((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt__call__s   N(t__name__t __module__t__doc__RRRRR R!tpropertyRR"R(R*R,R5(((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyRs!       tFactorcBsSeZdZed„Zd„Zd„Zd„Zd„Zdd„Z d„Z RS( sA categorical factor.cCsì|s+tt|ƒƒ|_|jjƒn<||_tt|ƒƒtt|ƒƒkrgtdƒ‚n||_||_||_|jr—|j}n/g|jD]}d|jt |ƒf^q¡}t j ||d|jd|j ƒdS(s& Factor is initialized with keys, representing all valid levels of the factor. If ordinal is False, keys can have repeats: set(keys) is what is used. If ordinal is True, the order is taken from the keys, and there should be no repeats. s=keys for ordinal Factor should be unique, in increasing orders(%s==%s)RRN( R+tsettkeystsorttlenRt_nameRtordinaltstrRRt get_columns(RRR<R@Rtkey((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR©s  $     /c Os-|j|j}x\trnt|ƒrjt|ttfƒrXtj|ƒ}|j|_n|||Ž}qPqWt|ƒ}|j rÆgt |ƒD]"}t |j j ||ƒƒ^q‘}tj|ƒSg}xM|j D]B}gt |ƒD]}t |||kƒ^qé}|j|ƒqÖWtj|ƒSdS(s7 Calling function for factor instance. N(RR?tTrueR.RRR#R/R>R@RRR<tindexR tarraytappend( RR2R3tvtntitcolRRC((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyRBÇs"    5 /cOs%|`|||Ž}|j|_|S(sf Return the keys of the factor, rather than the columns of the design matrix. (RRB(RR2R3R4((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pytvaluesâs cCs1t|ƒ}|j|jƒs-tdƒ‚ndS(sJ Verify that all values correspond to valid keys in self. sunknown keys in valuesN(R;tissubsetR<R(RRLts((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pytverifyís cCsEt|tƒr1|jdkr1t|d|jƒStj||ƒSdS(s¢ Formula(self) + Formula(other) When adding 'intercept' to a factor, this just returns Formula(self, namespace=self.namespace) R)RN(RRRR#RR((RR%((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR(õs c Cs|jƒ}|dkr!d}n6y|jj|ƒ}Wntk rVt|ƒ}nX|d„}tt|jƒƒƒ}|j|ƒ|jƒ}gt t|ƒƒD]"}d|||||f^q§}t |d|dd|j d|ƒ}|j |_ |S( s Return the 'main effect' columns of a factor, choosing an optional reference key. The reference key can be one of the keys of the Factor, or an integer, representing which column to remove. It defaults to 0. icSsmg}t|jdƒ}|j|ƒx7tt|ƒƒD]#}|j|||||ƒq9Wtj|ƒS(Ni(RtshapetpopRR>RGR RF(Rt referencetrvaluetkeepRJ((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pytmaineffect_funcs  !s%s-%sRRs %s:maineffectR N( R,RR<RERRRR>RQRRRR( RRRR,RURTt_Factor__namesRJt_namesR((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt main_effects"       ;   cCs‚|js<|jƒjd|jt|ƒfƒ}|ƒ|S|j|j}tjg|D]}||k^qYƒj tj ƒSdS(s Retrieve the column corresponding to key in a Formula. Parameters ---------- key : Factor key one of the Factor's keys Returns ------- ndarray corresponding to key, when evaluated in current namespace s(%s==%s)N( R@R,RERRARR?R RFtastypeR(RRCRJRHtvv((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt __getitem__+s ( N( R6R7R8tFalseRRBRLROR(RRXR[(((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR:¦s     'RcBs,eZdZddd„d„Zd„ZRS(s  A subclass of term that can be used to apply point transformations of another term, i.e. to take powers: >>> import numpy as np >>> from nipy.fixes.scipy.stats.models import formula >>> X = np.linspace(0,10,101) >>> x = formula.Term('X') >>> x.namespace={'X':X} >>> x2 = x**2 >>> print np.allclose(x()**2, x2()) True >>> x3 = formula.Quantitative('x2', func=x, transform=lambda x: x**2) >>> x3.namespace = x.namespace >>> print np.allclose(x()**2, x3()) True cCs|S(N((R ((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR TRcCs)||_tj||d|d|ƒdS(NRR(R RR(RRRRR ((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyRTs cOs|jtj|||ŽƒS(sz A quantitative is just like term, except there is an additional transformation: self.transform. (R RR5(RR2R3((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR5XsN(R6R7R8RRR5(((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR@sR#cBs¹eZdZd„Zd„Zd„ZeeeeƒZd„Ze d„Z d„Z d„Z d„Z d „Zed „Zd „Zd „Zd „Zed„Zd„Zd„ZRS(s8 A formula object for manipulating design matrices in regression models, essentially consisting of a list of term instances. The object supports addition and multiplication which correspond to concatenation and pairwise multiplication, respectively, of the columns of the two formulas. cCs-t|jtjƒr|jS|jp(tSdS(N(Rt_Formula__namespaceR RR(R((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyRkscCs ||_dS(N(R](RR((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR pRcCs |`dS(N(R](R((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR!qRcCs"|jƒ|_|jƒ|_dS(N(R,RWt termnamest _termnames(R((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt_terms_changedtscCsƒ||_t|tƒr6tjt|jƒƒ|_n?t|tƒrQ||_n$t|tƒro|g|_nt‚|jƒdS(s“ Create a formula from either: i. a `formula` object ii. a sequence of `term` instances iii. one `term` N( R]RR#R/R+ttermsRRR`(RttermlistR((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyRxs  cCs;g}x!|jD]}||jg7}qWddj|ƒS(sJ String representation of list of termnames of a formula. s s + (RaRtjoin(RRtterm((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR"sc Osd|kr|d}n |j}g}t}d}xÝ|jD]Ò}tj|ƒ}||_|||Ž}t} t|dƒr´|jdkr´t}t} |} |jd ƒq´n|j dkrð| rðd|j df|_ |j|ƒn| s|j|ƒn|d7}q>W|s:yt j |ƒ}WqôqôXnº|j ddƒ} |gkrº| dkrx|dj d} n|dj d} t jd| ft jƒ|| with only intercept in formula, keyword 'nrow' argument neededN(i(RR\RaR/R-RRDRGRtndimRPR t concatenatetgettonestfloat64RtIR1( RR2R3RtallvalsR)tiindexttR4t isintercepttinterceptindexReRI((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR5–sP       cCs·t|tƒsrt|tƒrMy!||}|j|jƒkSWqotSXq³t|tƒr³|j|jƒkSnAt|jƒdkr§|jd}|j|jƒkSt dƒ‚dS(sA Determine whether a given term is in a formula. iis$more than one term passed to hastermN( RR#RRR^R\RR>RaR(Rt query_termtquery((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pythastermÑs  cCsF|jƒ}||kr,|j|j|ƒStdt|ƒƒ‚dS(Nsformula has no such term: %s(R^RaREtKeyErrortrepr(RRRn((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR[ås  cCsx|j|ƒrN|jƒ}tƒ}x3|D]}|jj|ƒ||RRAtsortedtsplitRcRGRaR,RRR/RR$(RR%tnestedt selftermnamestothertermnamest n_term_namestJRaR^RJtjRtpiecest selfnamest othernamest_termR,R|R}RRNRR€tcselftcothertsumterms((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR*sT     $     $  cCs­d„}t|ƒ}|j|j}tg|D]}||jƒ|f^q/dd„ƒ}g|D]}|d^qc}t|ƒ}t|j|jƒr©|j|_n|S(s© Return a formula whose columns are the concatenation of the columns of self and other. terms in the formula are sorted alphabetically. cSst|tƒr|dS|SdS(Ni(RR+(R ((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt_delistlsRCcSs|dS(Ni((R ((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR tRi(R#RaRRR$R(RR%RRaRdR‰tpieceR'((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR(cs  7 cCs’t|ƒ}tj|jƒ}xU|jD]J}xAtt|ƒƒD]-}||j|jkrA|j|ƒPqAqAWq(Wt|ƒ}|j|_|S(s¹ Return a formula with all terms in other removed from self. If other contains term instances not in formula, this function does not raise an exception. (R#R/RaRR>RRQR(RR%RaRdRJR'((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt__sub__{s     (R6R7R8RR R!R9RR`RRR"R5RsR[R\RwR,R^RzR*R(R’(((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR#`s$       ;     H c CsC|dk r?tj|ƒ}||_tj|ƒ}||_n|dtƒd}|dtƒd}t|ƒt|ƒkrŒtdƒ‚ntt|ƒƒ}tt|ƒƒ}t||ƒ}gtt|ƒƒD]}||||f^qÒ} tt| ƒƒ} | |kr5||kr%|} n|} t| fSt dfSdS(só Is factor B nested within factor A or vice versa: a very crude test which depends on the namespace. If they are nested, returns (True, F) where F is the finest level of the relationship. Otherwise, returns (False, None) RLis2A() and B() should be sequences of the same lengthN( RR/RRDR>RR;tmaxRR\( tAtBRtatbtnAtnBRIRJtABtnABtF((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pytisnesteds(   3    icKstjd|fƒS(Ni(R Ri(Retextra((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt _intercept_fnµsR)Rs1 Intercept term in a formula. If intercept is the only term in the formula, then a keyword argument 'nrow' is needed. >>> from nipy.fixes.scipy.stats.models.formula import Formula, I >>> I() array(1.0) >>> I(nrow=5) array([ 1., 1., 1., 1., 1.]) >>> f=Formula(I) >>> f(nrow=5) array([1, 1, 1, 1, 1]) ic Cs t|ƒ}i}tj|ƒjdkrFtdt|ƒdƒ}nxt|D]l}tj|f|ƒ}|jdtj|jdƒf|_x't|jdƒD]}tj |dd…|fƒj|dd…|fjkr£tj tj tj |dd…|fƒ|dd…|fƒƒr£g|dd…|fD]}||^q?}|d} x0tt|ƒdƒD]} | || d9} qvW| |t |dd…|fƒ>> print interactions([Term(l) for l in ['a', 'b', 'c']]) >>> >>> print interactions([Term(l) for l in ['a', 'b', 'c']], order=list(range(5))) >>> iiN((R>R R0RPRRtindicestproductRtuniquetalltruetequalR=ttupleR+RR( Ratordertn_termsRLtoR tmRˆtllRHtiiRCR((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt interactionsÉs*  &;D- + cCs ||kS(N((tspace1tspace2((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyR$s(R8t collectionsRtstatsmodels.compat.pythonRRRRRR/tnumpyR t __docformat__RtobjectRR:RR#RRRŸRkR¬R$(((s:lib/python2.7/site-packages/statsmodels/sandbox/formula.pyt s"(  ‘š ÿ0 &  7