B ¤ŠãZP~ã@s6ddlmZdddgZddlZddlZddlmmZddl mm Z ddl m mZddlmZmZmZmZmZmZmZmZmZmZddlmZmZmZddlm Z m!Z!m"Z"m#Z#dd l$m%Z%m&Z&dd l'm(Z(d Z)Gd d „d eƒZ*Gdd„de*ƒZ+Gdd„de*ƒZ,Gdd„de*ƒZ-Gdd„deƒZ.Gdd„dee.ƒZ/Gdd„dej0ƒZ1e  2e1e.¡Gdd„dej0ƒZ3e  2e3e/¡Gdd„deƒZ4Gdd„dee4ƒZ5Gdd„dej0ƒZ6e  2e6e4¡Gdd „d ej0ƒZ7e  2e7e5¡Gd!d"„d"eƒZ8Gd#d$„d$ee8ƒZ9Gd%d&„d&ej0ƒZ:e  2e:e8¡Gd'd(„d(ej0ƒZ;e  2e;e9¡dS))é)ÚdivisionÚZeroInflatedPoissonÚZeroInflatedGeneralizedPoissonÚZeroInflatedNegativeBinomialPN) Ú DiscreteModelÚ CountModelÚPoissonÚLogitÚ CountResultsÚL1CountResultsÚProbitÚ_discrete_results_docsÚGeneralizedPoissonÚNegativeBinomialP)Ú zipoissonÚ zigenpoissonÚzinegbin)Ú approx_fprimeÚ approx_hessÚapprox_hess_csÚapprox_fprime_cs)Úresettable_cacheÚcache_readonly)ÚConvergenceWarninga exog_infl : array_like or None Explanatory variables for the binary inflation model, i.e. for mixing probability model. If None, then a constant is used. offset : array_like Offset is added to the linear prediction with coefficient equal to 1. exposure : array_like Log(exposure) is added to the linear prediction with coefficient equal to 1. inflation : string, 'logit' or 'probit' The model for the zero inflation, either Logit (default) or Probit c s®eZdZdejeejdœZd*‡fdd„ Zdd „Z d d „Z d+‡fdd„ Z e j je _d,‡fdd„ Z e j je _dd„Zdd„Zdd „Zd!d"„Zd#d$„Zd%d&„Zd-d(d)„Z‡ZS).ÚGenericZeroInflatedaQ Generiz Zero Inflated model for count data %(params)s %(extra_params)s Attributes ----------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. exog_infl: array A reference to the zero-inflated exogenous design. )ÚparamsÚ extra_paramsNÚlogitÚnonec  svtt|ƒj||f|||dœ|—Ž|dkrLd|_tj|j|jftjd|_n||_|j d|_t |j ƒdkrtd|_ n |j d|_ ||_ |dkr´t t |jj d¡|jƒ|_|j|_n:|dkrâtt |jj d¡|jƒ|_|j|_n td|ƒ‚||_|j|_t |jƒt |jƒkrtdƒ‚d d „|jjjDƒ} | t|jƒ|jdd…<tj|jtjd|_|j d d g¡d g|_ dS) N)ÚoffsetÚexposureÚmissingé)ÚdtyperrÚprobitz%inflation == %s, which is not handledz[exog and exog_infl have different number ofobservation. `missing` handling is not supportedcSsg|] }d|‘qS)z inflate_%s©)Ú.0Úir%r%ú?lib/python3.7/site-packages/statsmodels/discrete/count_model.pyú `sz0GenericZeroInflated.__init__..Ú exog_inflÚ inflation)!ÚsuperrÚ__init__Ú k_inflateÚnpÚonesÚsizeÚfloat64r*ÚshapeÚlenÚk_exogZinflr ÚzerosÚ model_inflÚ_hessian_logitÚ_hessian_inflater Ú_hessian_probitÚ TypeErrorr+Úk_extraÚexogÚ ValueErrorÚdataZ param_namesÚlistÚ exog_namesZasarrayZ _init_keysÚextendZ_null_drop_keys) ÚselfÚendogr=r*rr+r r!ÚkwargsZ infl_names)Ú __class__r%r(r-8s@      zGenericZeroInflated.__init__cCst | |¡¡S)a9 Loglikelihood of Generic Zero Inflated model Parameters ---------- params : array-like The parameters of the model. Returns ------- loglike : float The log-likelihood function of the model evaluated at `params`. See notes. Notes -------- .. math:: \ln L=\sum_{y_{i}=0}\ln(w_{i}+(1-w_{i})*P_{main\_model})+ \sum_{y_{i}>0}(\ln(1-w_{i})+L_{main\_model}) where P - pdf of main model, L - loglike function of main model. )r/ÚsumÚ loglikeobs)rCrr%r%r(ÚloglikegszGenericZeroInflated.loglikec CsÚ|d|j…}||jd…}|j}|j |¡}t |t t¡jdt t¡j¡}|j   |¡}t  |dk¡d}t  |¡d}tj |tj d} t ||d||t ||¡¡| |<t d||¡||| |<| S)ae Loglikelihood for observations of Generic Zero Inflated model Parameters ---------- params : array-like The parameters of the model. Returns ------- loglike : ndarray The log likelihood for each observation of the model evaluated at `params`. See Notes Notes -------- .. math:: \ln L=\ln(w_{i}+(1-w_{i})*P_{main\_model})+ \ln(1-w_{i})+L_{main\_model} where P - pdf of main model, L - loglike function of main model. for observations :math:`i=1,...,n` Nr"r)r#)r.rDr7Úpredictr/ÚclipÚfinfoÚfloatÚepsÚ model_mainrHÚnonzeroÚ zeros_liker2ÚlogÚexp) rCrÚ params_inflÚ params_mainÚyÚwÚllf_mainÚzero_idxÚ nonzero_idxÚllfr%r%r(rHs "  "zGenericZeroInflated.loglikeobsÚbfgsé#r"Ú nonrobustc  s¶|dkrBt|ddƒt|ddƒ} t | ¡dkr:| dkr:d} | ¡}|dkrRdd„}tt|ƒjf||||||dœ| —Ž} | || j¡} |  | ¡}|dkrši}|j f|d| d œ|—Ž|S) Nrrr r"cWs|S)Nr%)Úxr%r%r(Ú´sz)GenericZeroInflated.fit..)Ú start_paramsÚmaxiterÚdispÚmethodÚ full_outputÚcallbackT)Úcov_typeZuse_selfÚuse_t) Úgetattrr/r1Ú_get_start_paramsr,rÚfitÚ result_classZ_resultsÚresult_class_wrapperZ_get_robustcov_results)rCrardrbrercrfrgZcov_kwdsrhrErZmlefitZzipfitÚresult)rFr%r(rk©s$ zGenericZeroInflated.fitÚl1Údefined_by_methodrÚautoç{®Gáz„?ç-Cëâ6?縅ëQ¸ž?c  s>t |¡dkr0|dkr0|j|j} |t | ¡}|j|j}|jrdt |¡dkrd|d|j| …n|}|dkrât|ddƒt|ddƒ}t |¡dkr¢|dkr¢d}|jjf||||d|||| | | dœ | —Žj }t  t |j¡|¡}t t |ƒjf||||||||| | | dœ | —Ž}|dkr(|  ||¡}n td|ƒ‚| |¡S)Nr"rrr ) rardrbrercrfÚalphaÚ trim_modeÚ auto_trim_tolÚ size_trim_tolÚqc_tol)roZ l1_cvxopt_cpz+argument method == %s, which is not handled)r/r1r5r.r0r<rirOÚfit_regularizedrÚappendr,rÚresult_class_regr;Úresult_class_reg_wrapper)rCrardrbrercrfrurvrwrxryrEZk_paramsZextraZalpha_prZcntfitZ discretefit)rFr%r(rzÇs4  &   z#GenericZeroInflated.fit_regularizedcCsº|d|j…}||jd…}|j}|j |¡}t |t t¡jdt t¡j¡}|j   |¡}|j   |¡}|  |¡}t  |dk¡d} t  |¡d} |j  |¡} tj |jjd|jftjd} tj|jtjd} || jd|| t || ¡j| | dd…f<|| | | dd…f<|jdkr”|j| j|| d|| dt || ¡t || ¡j| | dd…f<|j| j|| j | | dd…f<n|jdkr¬t||j ƒSt | | f¡S)aŠ Generic Zero Inflated model score (gradient) vector of the log-likelihood Parameters ---------- params : array-like The parameters of the model Returns ------- score : ndarray, 1-D The score vector of the model, i.e. the first derivative of the loglikelihood function, evaluated at `params` Nr"r)r#rr$)r.rDr7rJr/rKrLrMrNrOÚ score_obsrHrPr6r=r3r5r2rQr*ÚTrSr+rZhstack)rCrrTrUrVrWÚ score_mainrXr[rYrZÚmuZdldpZdldwr%r%r(r~ís0 "    * 0   zGenericZeroInflated.score_obscCs| |¡ d¡S)Nr)r~rG)rCrr%r%r(ÚscoreszGenericZeroInflated.scorecCsdS)Nr%)rCrr%r%r(Ú _hessian_main sz!GenericZeroInflated._hessian_mainc Cs.|d|j…}||jd…}|j}|j |¡}t |t t¡jdt t¡j¡}|j   |¡}|j   |¡}|  |¡}t  |dk¡d} t  |¡d} t  |j|j|jf¡} t |¡} xþt|jƒD]ð} xêt| ddƒD]Ú}|j| | f|j| |f|| d|| dt || ¡dd|| t || ¡|| || ddt || ¡d| | d ¡|j| | f|j| |f|| d||  ¡| | |f<qÒWqÀWxtt|jƒD]f} x^t|jƒD]P}|| |f|| d|| |j| | f| |  ¡ | | ||jf<qÐWqÀW| S)Nr"réÿÿÿÿé)r.rDr7rJr/rKrLrMrNrOr~rHrPr6r5rSÚranger*rG)rCrrTrUrVrWr€rXr[rYrZÚhess_arrÚpmfr'Újr%r%r(r8#s, "    lFVz"GenericZeroInflated._hessian_logitcCsdS)Nr%)rCrr%r%r(r:Jsz#GenericZeroInflated._hessian_probitcCs | |¡}| |¡}|dks$|dkr0t||jƒS|j|j}t ||f¡}||d|j…dd…f<|||jd…|jd…f<tj|j|jdd}|j |||<|S)a‹ Generic Zero Inflated model Hessian matrix of the loglikelihood Parameters ---------- params : array-like The parameters of the model Returns ------- hess : ndarray, (k_vars, k_vars) The Hessian, second derivative of loglikelihood function, evaluated at `params` Notes ----- Nr")Úk) rƒr9rrIr5r.r/r6Z triu_indicesr)rCrZ hess_arr_mainZ hess_arr_inflZdimr‡Ztri_idxr%r%r(ÚhessianMs    zGenericZeroInflated.hessianÚmeancCs"|dkr|j}|dkr|j}|dkr2t|ddƒ}n t |¡}|dkrHd}|d|j…}||jd…}d|j ||¡} t ||d|jj d…¡||} |j j} |j j } t|j ddgƒ} t|j ddgƒ}||j _t  |j d¡|j _ ||j _ ||j _|j  |¡}| |j _| |j _ t| ƒdkr6| ddkr6|j ` n| |j _ t|ƒdkrb|ddkrb|j `n||j _d| | t |¡}|dkr˜| t | ¡S|dkr¬t | ¡S|d krº| S|d krÚ| t | ¡d|S|d krè|S|d krö| S|d kr| |||||¡Std|ƒ‚dS)aH Predict response variable of a count model given exogenous variables. Parameters ---------- params : array-like The parameters of the model exog : array, optional A reference to the exogenous design. If not assigned, will be used exog from fitting. exog_infl : array, optional A reference to the zero-inflated exogenous design. If not assigned, will be used exog from fitting. offset : array, optional Offset is added to the linear prediction with coefficient equal to 1. exposure : array, optional Log(exposure) is added to the linear prediction with coefficient equal to 1. If exposure is specified, then it will be logged by the method. The user does not need to log it first. which : string, optional Define values that will be predicted. 'mean', 'mean-main', 'linear', 'mean-nonzero', 'prob-zero, 'prob', 'prob-main' Default is 'mean'. Notes ----- Nr rr"rZnorŒz mean-mainÚlinearz mean-nonzeroz prob-zeroz prob-mainZprobzwhich = %s is not available)r=r*rir/rRr.r7rJÚdotr3rOrDr6rr rHr4rSÚ _predict_probr>)rCrr=r*r rÚwhichrTrUZ prob_mainZlin_predZtmp_exogZ tmp_endogZ tmp_offsetZ tmp_exposurer[Z prob_zeror%r%r(rJqs\ $         zGenericZeroInflated.predict)NNrNr) Nr\r]r"r"Nr^NN) Nrorpr"r"Nrrqrrrsrt)NNNNrŒ)Ú__name__Ú __module__Ú __qualname__ÚbaseÚ_model_params_docÚ_doc_zi_paramsÚ_missing_param_docÚ__doc__r-rIrHrkrrzr~r‚rƒr8r:r‹rJÚ __classcell__r%r%)rFr(r&s..* ! 0'$rcsLeZdZdejeejdœZd‡fdd„ Zdd „Z d d „Z d d „Z ‡Z S)raQ Poisson Zero Inflated model for count data %(params)s %(extra_params)s Attributes ----------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. exog_infl: array A reference to the zero-inflated exogenous design. )rrNrrc  s^tt|ƒj||f|||||dœ|—Žt|j|j||d|_t|_t |_ t |_ t |_t|_dS)N)rr+r*r r!)rr )r,rr-rrDr=rOrÚ distributionÚZeroInflatedPoissonResultsrlÚ!ZeroInflatedPoissonResultsWrapperrmÚL1ZeroInflatedPoissonResultsr|Ú#L1ZeroInflatedPoissonResultsWrapperr}) rCrDr=r*rr r+r!rE)rFr%r(r-âs  zZeroInflatedPoisson.__init__c Csv|d|j…}||jd…}|j}|j |¡}t |t t¡jdt t¡j¡}|  |¡}t  |dk¡d}t  |¡d}|j  |¡} t  |j |j f¡} d||t | |¡d} xºt|j ƒD]¬} x¦t| ddƒD]–} |j|| f|j|| f| |||dd| ||| |t | |¡| d ¡| ||j|| f|j|| f ¡| | | f<qÔWqÂW| S)Nr"rr„r…)r.rDr7rJr/rKrLrMrNr‚rPrOr6r5rSr†r=rG)rCrrTrUrVrWr‚rYrZrr‡Zcoeffr'r‰r%r%r(rƒñs$ "  4$&z!ZeroInflatedPoisson._hessian_mainc Csì|d|j…}||jd…}t t dt |j¡d¡¡}t|jƒdkrnd} t |j  ||¡¡dd…df} nd} |j  ||¡dd…df} t  | t  t ¡j dt  t ¡j ¡} |jj |||ddd…df} |j || | ¡} | rè| dS| S)Nrr"r…TF)r)r.r/Ú atleast_2dÚarangeÚmaxrDr4r3r7rJrKrLrMrNrOršrˆ) rCrr=r*r rrTrUÚcountsÚ transformrWrrnr%r%r(rs" z!ZeroInflatedPoisson._predict_probcCs.|jjdddj}t t |j¡d|¡}|S)NrÚnm)rcrdgš™™™™™¹?)rOrkrr/r{r0r.)rCrar%r%r(rj"sz%ZeroInflatedPoisson._get_start_params)NNNrr) r‘r’r“r”r•r–r—r˜r-rƒrrjr™r%r%)rFr(rÐscsTeZdZdejedejdœZd‡fdd „ Z‡fd d „Z d d „Z dd„Z ‡Z S)ra¢ Zero Inflated Generalized Poisson model for count data %(params)s %(extra_params)s Attributes ---------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. exog_infl: array A reference to the zero-inflated exogenous design. p: scalar P denotes parametrizations for ZIGP regression. zŽp : float dispersion power parameter for the GeneralizedPoisson model. p=1 for ZIGP-1 and p=2 for ZIGP-2. Default is p=2 )rrNrr…rc  sˆtt|ƒj||f|||||dœ| —Žt|j|j|||d|_t|_|j d7_ |j d7_ |j   d¡t |_t|_t|_t|_dS)N)rr+r*r r!)rr Úpr"ru)r,rr-rrDr=rOrršr5r<rAr{Ú%ZeroInflatedGeneralizedPoissonResultsrlÚ,ZeroInflatedGeneralizedPoissonResultsWrapperrmÚ'L1ZeroInflatedGeneralizedPoissonResultsr|Ú.L1ZeroInflatedGeneralizedPoissonResultsWrapperr}) rCrDr=r*rr r+r¥r!rE)rFr%r(r-@s   z'ZeroInflatedGeneralizedPoisson.__init__cs"tt|ƒ ¡}|jjd|d<|S)Nr"r¥)r,rÚ_get_init_kwdsrOÚparameterization)rCÚkwds)rFr%r(rªSsz-ZeroInflatedGeneralizedPoisson._get_init_kwdscCsð|d|j…}||jd…}|jj}t t dt |j¡d¡¡} t|j ƒdkrvd} t |j   ||¡¡dd…df} nd} |j   ||¡dd…df} t  dd¡| | dk<|jj ||||ddd…df} |j  | | |d|| ¡} | rì| dS| S) Nrr"r…TFgð?)r rr„)r.rOr«r/rŸr r¡rDr4r3r7rJZ nextafterršrˆ)rCrr=r*r rrTrUr¥r¢r£rWrrnr%r%r(rXs z,ZeroInflatedGeneralizedPoisson._predict_probc CsPt ¡2tjdtdt|j|j|jdjddj }WdQRXt   |d¡}|S)NÚignore)Úcategory)r*r)rcgš™™™™™¹?) ÚwarningsÚcatch_warningsÚ simplefilterrrrDr=r*rkrr/r{)rCrar%r%r(rjms    z0ZeroInflatedGeneralizedPoisson._get_start_params)NNNrr…r) r‘r’r“r”r•r–r—r˜r-rªrrjr™r%r%)rFr(r(s csTeZdZdejedejdœZd‡fdd „ Z‡fd d „Z d d „Z dd„Z ‡Z S)raã Zero Inflated Generalized Negative Binomial model for count data %(params)s %(extra_params)s Attributes ----------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. exog_infl: array A reference to the zero-inflated exogenous design. p: scalar P denotes parametrizations for ZINB regression. p=1 for ZINB-1 and p=2 for ZINB-2. Default is p=2 zp : float dispersion power parameter for the NegativeBinomialP model. p=1 for ZINB-1 and p=2 for ZINM-2. Default is p=2 )rrNrr…rc  sˆtt|ƒj||f|||||dœ| —Žt|j|j|||d|_t|_|j d7_ |j d7_ |j   d¡t |_t|_t|_t|_dS)N)rr+r*r r!)rr r¥r"ru)r,rr-rrDr=rOrršr5r<rAr{Ú#ZeroInflatedNegativeBinomialResultsrlÚ*ZeroInflatedNegativeBinomialResultsWrapperrmÚ%L1ZeroInflatedNegativeBinomialResultsr|Ú,L1ZeroInflatedNegativeBinomialResultsWrapperr}) rCrDr=r*rr r+r¥r!rE)rFr%r(r-s   z&ZeroInflatedNegativeBinomialP.__init__cstt|ƒ ¡}|jj|d<|S)Nr¥)r,rrªrOr«)rCr¬)rFr%r(rª¢s z,ZeroInflatedNegativeBinomialP._get_init_kwdscCsø|d|j…}||jd…}|jj}t dt |j¡d¡} t|jƒdkrpd} t  |j   ||¡¡dd…df} nd} |j   ||¡dd…df} t  | t  t¡jdt  t¡j¡} |jj ||||ddd…df} |j | | |d|| ¡} | rô| dS| S)Nrr"r…TF)r rr„)r.rOr«r/r r¡rDr4r3rŸr7rJrKrLrMrNršrˆ)rCrr=r*r rrTrUr¥r¢r£rWrrnr%r%r(r§s" z+ZeroInflatedNegativeBinomialP._predict_probc CsLt ¡&tjdtd|jjdddj}WdQRXt t  |j ¡|¡}|S)Nr­)r®rr¤)rcrd) r¯r°r±rrOrkrr/r{r6r.)rCrar%r%r(rj¼s  z/ZeroInflatedNegativeBinomialP._get_start_params)NNNrr…r) r‘r’r“r”r•r–r—r˜r-rªrrjr™r%r%)rFr(rvs c@s0eZdZedddœZedd„ƒZd d d „ZdS) r›z)A results class for Zero Inflated PoissonÚ)Úone_line_descriptionÚ extra_attrcCs<|jdd}d| ¡t |jdd¡}d|t |¡S)Nr)rr")rJr/rS)rCrrWr%r%r(Ú_dispersion_factorÉs z-ZeroInflatedPoissonResults._dispersion_factorÚoverallÚdydxNFcCs tdƒ‚dS)zhGet marginal effects of the fitted model. Not yet implemented for Zero Inflated Models z¬ yet implemented for zero inflationN)ÚNotImplementedError)rCÚatrdÚatexogÚdummyÚcountr%r%r(Ú get_margeffÏsz&ZeroInflatedPoissonResults.get_margeff)rºr»NFF)r‘r’r“r r˜rr¹rÁr%r%r%r(r›Äs   r›c@s eZdZdS)rN)r‘r’r“r%r%r%r(rØsrc@s eZdZdS)rœN)r‘r’r“r%r%r%r(rœÜsrœc@s eZdZdS)ržN)r‘r’r“r%r%r%r(ržâsržc@s0eZdZedddœZedd„ƒZd d d „ZdS) r¦z5A results class for Zero Inflated Generalized Poissonr¶)r·r¸cCs^|jjj}|j|jjd…d}t |jdd¡}d| ¡|}d|||d||S)Nr„r)rr"r…)ÚmodelrOr«rr.r/rSrJ)rCr¥rurrWr%r%r(r¹ís  z8ZeroInflatedGeneralizedPoissonResults._dispersion_factorrºr»NFcCs tdƒ‚dS)zhGet marginal effects of the fitted model. Not yet implemented for Zero Inflated Models z¬ yet implemented for zero inflationN)r¼)rCr½rdr¾r¿rÀr%r%r(rÁõsz1ZeroInflatedGeneralizedPoissonResults.get_margeff)rºr»NFF)r‘r’r“r r˜rr¹rÁr%r%r%r(r¦ès   r¦c@s eZdZdS)r¨N)r‘r’r“r%r%r%r(r¨þsr¨c@s eZdZdS)r§N)r‘r’r“r%r%r%r(r§sr§c@s eZdZdS)r©N)r‘r’r“r%r%r%r(r© sr©c@s0eZdZedddœZedd„ƒZd d d „ZdS) r²z?A results class for Zero Inflated Genaralized Negative Binomialr¶)r·r¸cCs^|jjj}|j|jjd…d}t |jdd¡}d| ¡|}d|||d||S)Nr„r)rr")rÂrOr«rr.r/rSrJ)rCr¥rurrWr%r%r(r¹s  z6ZeroInflatedNegativeBinomialResults._dispersion_factorrºr»NFcCs tdƒ‚dS)zhGet marginal effects of the fitted model. Not yet implemented for Zero Inflated Models z¬ yet implemented for zero inflationN)r¼)rCr½rdr¾r¿rÀr%r%r(rÁsz/ZeroInflatedNegativeBinomialResults.get_margeff)rºr»NFF)r‘r’r“r r˜rr¹rÁr%r%r%r(r²s   r²c@s eZdZdS)r´N)r‘r’r“r%r%r%r(r´'sr´c@s eZdZdS)r³N)r‘r’r“r%r%r%r(r³,sr³c@s eZdZdS)rµN)r‘r’r“r%r%r%r(rµ3srµ)sd 0 -XNN