M/PhgdZddlZddlZddlmcmZddlmcmZ ddl m cm Z ddlmZmZddlmZmZmZmZmZmZmZmZddlmZddlmZddlmZddl m!Z!Gd d eZ"Gd d e"Z#Gd de"Z$Gdde"Z%GddeZ&Gdde&Z'Gdde&Z(Gdde&Z)Gdde&Z*GddeZ+GddeZ,Gdd e,Z-Gd!d"e,Z.Gd#d$ee,Z/Gd%d&e j0Z1e j2e1e,Gd'd(e j0Z3e j2e3e/Gd)d*eZ4Gd+d,ee4Z5Gd-d.e j0Z6e j2e6e4Gd/d0e j0Z7e j2e7e5dS)1)TruncatedLFPoissonTruncatedLFNegativeBinomialPHurdleCountModelN)truncatedpoissontruncatednegbin) DiscreteModel CountModel CountResultsL1CountResultsPoissonNegativeBinomialPGeneralizedPoisson_discrete_results_docs) approx_hess)cache_readonly)ConvergenceWarning)deepcopyceZdZdejdejzdzZ dfd ZdZd Z d Z d Z dfd Z e j je _ dfd Ze jje_dZ ddZxZS)TruncatedLFGenerica Generic Truncated model for count data .. versionadded:: 0.14.0 %(params)s %(extra_params)s Attributes ---------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. truncation : int, optional Truncation parameter specify truncation point out of the support of the distribution. pmf(k) = 0 for k <= truncation 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. params extra_paramsrNnonec Ttj||f|||d||j|k}|j||_|j||_||j||_||j||_||_||_|j dgg|_ dS)Noffsetexposuremissing truncation) super__init__endogexogrrtruncr _init_keysextend_null_drop_keys) selfr#r$r rrrkwargsmask __class__s d/var/www/html/test/jupyter/venv/lib/python3.11/site-packages/statsmodels/discrete/truncated_model.pyr"zTruncatedLFGeneric.__init__9s           zJ&IdO Z%  +d+DK   M$/DM $  ~...!cPtj||S)a` Loglikelihood of Generic Truncated 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 ----- npsum loglikeobsr)rs r-loglikezTruncatedLFGeneric.loglikeQ &vdoof--...r.c|j|}|jdz}||dt j|d}t j|tj }|dk}tj ||<|dk}t j d||z ||<||z }|S)a Loglikelihood for observations of Generic Truncated model Parameters ---------- params : array-like The parameters of the model. Returns ------- loglike : ndarray (nobs,) The log likelihood for each observation of the model evaluated at `params`. See Notes Notes -----  prob-basewhichy_values) model_mainr3r%predictr1aranger2 full_likeinfnanlog)r)rllf_mainytpmf log_1_m_pmflocllfs r-r3zTruncatedLFGeneric.loglikeobsfs&?--f55 Z!^ll + " ???Bs2ww l300 Ag6 CAg6!c#h,// C$ r.c z|j|}tj|jtj}tj|tj}t |jdzD]}|jtj |j|z|j t|ddt|dd}tj | |}|||j|zjz }||z }||jd|z z jz}|S) Generic Truncated 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` )dtyper8rNr)rr)r> score_obsr1 zeros_liker#float64ranger%r, ones_liker$getattrexpr3T) r)r score_mainrG score_truncimodelpmf_idparamss r-rNzTruncatedLFGeneric.score_obss_..v66 mDJbj999mJbjAAA tzA~&&  AO-- TZ((1, tXt44 z488 .E F5++F3344E EOOF335=@ @K 5LCC S 9<<r.cR||dS)rLrrNr2r4s r-scorezTruncatedLFGeneric.score$~~f%%))!,,,r.bfgs#r8 nonrobustc |t|ddt|ddz} tj| dkr| dkrd} |j|j|j| } tj5tj dt| dj }dddn #1swxYwY|j dz|jz} |jjd| z |_t#j d |||||d d | }|||j}||}|i}|jd |d | d ||S)Nrrrr8rignorecategorydispc|SNxs r-z(TruncatedLFGeneric.fit..qr. start_paramsmethodmaxiterri full_outputcallbackTcov_typeuse_selfuse_trl)rSr1sizer>r,r#r$warningscatch_warnings simplefilterrfitrdf_modelk_extrashapedf_residr! result_class_resultsresult_class_wrapper_get_robustcov_results)r)rrrsrtrurirvrxcov_kwdsrzr*rrYk_paramsmlefitzipfitresultr,s r-rzTruncatedLFGeneric.fits  T8Q//'$ A2N2NNFwv!##! O--dj$)5;.==E(** 8 8%h9KLLLL$yyay007  8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 =1$t|3 (+h6 %# [ ""499**622  H%% Nx/35 N NDL N N N 7CC Cl1defined_by_methodauto{Gz?-C6?Q?c  tj|dkr/|dkr)|jjd} |tj| z}|}|t |ddt |ddz}tj|dkr|dkrd}|j|j|j|}|j d ||||d|||| | | d | j }tt|j d ||||||||| | | d | }|dvr| ||}ntd|z||S Nr8rrrrd) rrrsrtrurirvalpha trim_mode auto_trim_tol size_trim_tolqc_tol)r l1_cvxopt_cpz+argument method == %s, which is not handledrlr1r{r$ronesrSr>r,r#fit_regularizedrr!r result_class_reg TypeErrorresult_class_reg_wrapperr)rrrsrtrurirvrrrrrr*ralpha_prrYcntfit discretefitr,s r-rz"TruncatedLFGeneric.fit_regularized 75>>Q  5A::yq)HBGH---E  T8Q//'$ A2N2NNFwv!##! O--dj$)5;.==E050F)&''a(++F FF ?E FF GM  9z4((8F)&''dXy +F FF?E FF + + +//f==KKAFJLL L,,[999r.c,t||jS)a Generic Truncated 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 ----- rr5r4s r-hessianzTruncatedLFGeneric.hessian$64<000r.meanc||||\}}}tj||d|jd}||z|z}|dkrtj|} |jdkr |j| |} | | z S|jdkr| S|jdkrtjtj d|jdz} |j ||tj||d| } | d} tj |jdz| z d}| |z d| z z }|Std |d kr|S|d krtj|S|dkr|tj|} nAtjtj dtj |jdz} tj|dddf} |jdkr"|j| | |j} nP|jdkr6|jj}|j| | |d||j} ntd | S|d krz|tj|} n/tj dtj |jdz} |j ||tj||d| } | S|dkr2tj|} tjtj d|jdz} |j ||tj||d| } | d} tj |jdz| z d}| |z d| z z }tj |jdzdz| z d}|j| |}| dz|z|z d| z z }||dzz }|Std|z)a Predict response variable or other statistic given exogenous variables. Parameters ---------- params : array_like The parameters of the model. exog : ndarray, optional Explanatory variables for the main count model. If ``exog`` is None, then the data from the model will be used. offset : ndarray, optional Offset is added to the linear predictor of the mean function with coefficient equal to 1. Default is zero if exog is not None, and the model offset if exog is None. exposure : ndarray, optional Log(exposure) is added to the linear predictor 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. Default is one if exog is is not None, and it is the model exposure if exog is None. which : str (optional) Statitistic to predict. Default is 'mean'. - 'mean' : the conditional expectation of endog E(y | x) - 'mean-main' : mean parameter of truncated count model. Note, this is not the mean of the truncated distribution. - 'linear' : the linear predictor of the truncated count model. - 'var' : returns the estimated variance of endog implied by the model. - 'prob-trunc' : probability of truncation. This is the probability of observing a zero count implied by the truncation model. - 'prob' : probabilities of each count from 0 to max(endog), or for y_values if those are provided. This is a multivariate return (2-dim when predicting for several observations). The probabilities in the truncated region are zero. - 'prob-base' : probabilities for untruncated base distribution. The probabilities are for each count from 0 to max(endog), or for y_values if those are provided. This is a multivariate return (2-dim when predicting for several observations). y_values : array_like Values of the random variable endog at which pmf is evaluated. Only used if ``which="prob"`` Returns ------- predicted values Notes ----- If exposure is specified, then it will be logged by the method. The user does not need to log it first. r$rrNr8rrr=prob)r$rrr;r<zunsupported self.truncationlinear mean-mainzk_extra is not 0 or 1r9varz argument which == %s not handled)_get_predict_arraysr1dotrrTr r> _prob_nonzero atleast_2dr@r?r2 ValueErrormaxr#r model_distrGr%parameterizationasarray_var)r)rr$rrr;r<fittedlinpredmu prob_maincountsprobs prob_tregion mean_tregionrp mnc2_tregionvmmnc2vs r-r?zTruncatedLFGeneric.predict*sVt"&!9!9":""fh f^djm^4558#f, F??B!## O99"fEE I~%B&& 1$$ryDOa4G'H'HII//x0@0@!&0BB %yy|| " $/A*= > > FKKANN \)a,.>?  !>??? h  N k ! !6'?? " f__#x00ryBF4:4F4Fq4H'I'IJJD)B|q  ++FB CC""O4++FBr ,-tz;;!!8999L k ! !#H--1bfTZ&8&8&:;;O++TBF8,<,<Vf,>>EL e^^B]29Q!0C#D#DEEFO++TBF8,<,<Vf,>>E!99Q<(.""""""0///*,,,\!!!F---"=?,07;######J #+CK-1IMIM #:#:#:#:#:#:J,;CO111(@D'+G<G<G<G<G<G<G<GrrTruncatedLFPoissonResultsr TruncatedLFGenericResultsWrapperrL1TruncatedLFGenericResultsr"L1TruncatedLFGenericResultsWrapperr) r)r#r$rrr rr*r,s r-r"zTruncatedLFPoisson.__init__s   !      "$*di+24T+J+J)0x)F)F$$$+5$D! ;(J%%%r.c`dtj| z }||z }|d|z |dzzz }||fS)Predict mean and variance of zero-truncated distribution. experimental api, will likely be replaced by other methods Parameters ---------- params : array_like The model parameters. This is only used to extract extra params like dispersion parameter. mu : array_like Array of mean predictions for main model. Returns ------- Predicted conditional variance. r8r)r1rT)r)rrwmvar_s r-_predict_mom_trunc0z&TruncatedLFPoisson._predict_mom_trunc0s?"_ FAEQT>!$wr.)NNrr rrrrrrrr"rrrs@r-rrs"+   !  " # ###G6;?'-KKKKKK,r.rcTeZdZdejdejzdzZ d fd Zd ZxZ S) ra Truncated Generalized Negative Binomial model for count data .. versionadded:: 0.14.0 %(params)s %(extra_params)s Attributes ---------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. truncation : int, optional Truncation parameter specify truncation point out of the support of the distribution. pmf(k) = 0 for k <= truncation rrNrrrc tj||f||||d|t|j|jt |ddt |dd||_|jj|_|j |jj|j dt|_ t|_ t|_t |_t$|_dSNrrr)rrr)r!r"r r#r$rSr>r exog_namesr'rr TruncatedNegativeBinomialResultsrrrrrrr r)r#r$rrr rrr*r,s r-r"z%TruncatedLFNegativeBinomialP.__init__s   !      , J IT:t444400 .  t94<-..IJJJ)<$D! ;(J%%%r.c|d}|jj}d|||dz zzzd|z z}d|z }||z }|d|||dz zzzz}|dz|z|z } | |dzz } || fS)rr=r8r)r>r) r)rrrr prob_zerorrrrrs r-rz0TruncatedLFNegativeBinomialP._predict_mom_trunc04s&r  O ,ac**cEk: M F 1urAaCy(( )A aad{$wr.NNrrrrrs@r-rrs"+   !  " # ###G6;?,2KKKKKK6r.rcNeZdZdejdejzdzZ d fd ZxZS) TruncatedLFGeneralizedPoissona Truncated Generalized Poisson model for count data .. versionadded:: 0.14.0 %(params)s %(extra_params)s Attributes ---------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. truncation : int, optional Truncation parameter specify truncation point out of the support of the distribution. pmf(k) = 0 for k <= truncation rrNrrrc tj||f||||d|t|j|jt |ddt |dd||_|jj|_|j |jj|j dd|_ t|_ t|_t|_t"|_dSr)r!r"rr#r$rSr>rrr'rrrrrrrrrrs r-r"z&TruncatedLFGeneralizedPoisson.__init__os   !      - J IT:t444400 .  t94<-..IJJJ<$D! ;(J%%%r.r rrrrrrrr"rrs@r-rrSs"+   !  " # ###G6;?,2KKKKKKKKKKr.rceZdZdejdejzdzZ dfd ZdZdZ d Z d Z dfd Z e j je _ dfd Ze jje_dZxZS)_RCensoredGenerica Generic right Censored 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. rrNrc tj|dkd|_tj|d|_t j||f|||d|dS)Nrr)r1nonzerozero_idx nonzero_idxr!r"r)r#r$rrrr*r,s r-r"z_RCensoredGeneric.__init__s~ 5A:..q1 :e,,Q/             r.cPtj||S)a_ Loglikelihood of Generic Censored 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 ----- r0r4s r-r5z_RCensoredGeneric.logliker6r.c |j|}tj||jtjdtj||jz f}|S)a Loglikelihood for observations of Generic Censored model Parameters ---------- params : array-like The parameters of the model. Returns ------- loglike : ndarray (nobs,) The log likelihood for each observation of the model evaluated at `params`. See Notes Notes ----- r8)r>r3r1 concatenaterrDrTr)r)rrErJs r-r3z_RCensoredGeneric.loglikeobssd&?--f55n dm $ VAx(89::: ; ; =  r.c b|j|}|j|}tj||j||jjtj||j zdtj||jz z jf}|S) Generic Censored 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` r8) r>rNr3r1rrrrUrT)r)rrVrEr^s r-rNz_RCensoredGeneric.score_obss_..v66 ?--f55 t} % ( ) +fXd./00 01"&$"23444678   r.cR||dS)rrr]r4s r-r^z_RCensoredGeneric.scorer_r.r`rar8rbc n|t|ddt|ddz} tj| dkr| dkrd} |j|j|j| } tj5tj dt| dj }dddn #1swxYwYtj d |||||d d | } ||| j}||}|i}|jd |d | d ||S)Nrrrr8rdrerfrhc|Srkrlrms r-roz'_RCensoredGeneric.fit..rpr.rqTrwrl)rSr1r{r>r,r#r$r|r}r~rrrr!rrrr)r)rrrsrtrurirvrxrrzr*rrYrrrr,s r-rz_RCensoredGeneric.fit s  T8Q//'$ A2N2NNFwv!##! O--dj$)5;.==E(** 8 8%h9KLLLL$yyay007  8 8 8 8 8 8 8 8 8 8 8 8 8 8 8%# [ ""499**622  H%% Nx/35 N NDL N N N rrrrrrrrc  tj|dkr/|dkr)|jjd} |tj| z}|}|t |ddt |ddz}tj|dkr|dkrd}|j|j|j|}|j d ||||d|||| | | d | j }tt|j d ||||||||| | | d | }|dvr| ||}ntd|z||Srrrs r-rz!_RCensoredGeneric.fit_regularized.rr.c,t||jS)a Generic Censored 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 ----- rr4s r-rz_RCensoredGeneric.hessianUrr.NNrrr)rrrrrrrr"r5r3rNr^rrrrrrs@r-rrs6 +   !  " # ##G,;?      ///*86---"=?,07;@ #+CK-1IMIM #:#:#:#:#:#:J,;CO1111111r.rcNeZdZdejdejzdzZ dfd ZxZS)_RCensoredPoissonz Censored 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. rrNrc tj||f|||d|ttj|j|j|_d|_t|_ t|_ t|_t|_dSNr)r!r"r r1rOr#r$r>rTruncatedLFGenericResultsrrrrrrrrs r-r"z_RCensoredPoisson.__init__s          ""- ";";TYGG5$D! ;(J%%%r.rrrs@r-rrjs +   !  " # ##G,,0(.KKKKKKKKKKr.rcNeZdZdejdejzdzZ dfd ZxZS) _RCensoredGeneralizedPoissona Censored 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. rrNrrc tj||f|||d|ttj|j|j|_d|_t|_ t|_ t|_t|_dSr)r!r"rr1rOr#r$r>rrrrrrrrr r)r#r$rrrrr*r,s r-r"z%_RCensoredGeneralizedPoisson.__init__s 4 ' & ' '% ' ' '- M$* % %ty225$D! ;(J%%%r.NrNrrrs@r-r r s +   !  " # ##G,45(. K K K K K K K K K Kr.r cNeZdZdejdejzdzZ dfd ZxZS) _RCensoredNegativeBinomialPa Censored 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. rrNrrc tj||f|||d|ttj|j|j||_d|_t|_ t|_ t|_t|_dS)Nrr)r!r"r r1rOr#r$r>rrrrrrrrrr s r-r"z$_RCensoredNegativeBinomialP.__init__s           ,BM$*,E,E,0I./...5$D! ;(J%%%r.r rrs@r-rrs +   !  " # ##G,45(.KKKKKKKKKKr.rcZeZdZdejdejzdzZeedddffd Z dZ xZ S) _RCensoredz Censored 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. rrNrc tj||f|||d||tj|j|j|_||_|jjx|_} | dkr-|j |jj | dt|_ t|_t|_t"|_dS)Nrr)r!r"r1rOr#r$r>rrrr'rrrrrrrr) r)r#r$rY distributionrrrr*rr,s r-r"z_RCensored.__init__s             % dj 9 949EE&!%!88 w Q;; O " "4?#=whii#H I I I5$D! ;(J%%%r.c<|j||}|S)zrProbability that count is not zero internal use in Censored model, will be refactored or removed )r>r)r)rrprob_nzs r-rz_RCensored._prob_nonzeros ///F;;r.) rrrrrrrr rr"rrrs@r-rrs +   !  " # ##G,+2.tKKKKKK.r.rceZdZdejdejzdzZ dfd Zd Zd Z ddZ e j je _ ddZ xZ S)ra Hurdle model for count data .. versionadded:: 0.14.0 %(params)s %(extra_params)s Attributes ---------- endog : array A reference to the endogenous response variable exog : array A reference to the exogenous design. dist : string Log-likelihood type of count model family. 'poisson' or 'negbin' zerodist : string Log-likelihood type of zero hurdle model family. 'poisson', 'negbin' p : scalar Define parameterization for count model. Used when dist='negbin'. pzero : scalar Define parameterization parameter zero hurdle model family. Used when zerodist='negbin'. aJoffset : 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. Notes ----- The parameters in the NegativeBinomial zero model are not identified if the predicted mean is constant. If there is no or only little variation in the predicted mean, then convergence might fail, hessian might not be invertible or parameter estimates will have large standard errors. References ---------- not yet rNpoissonrrc  ||d} t| tj||f||| d| d|_d|_|||||t |_t|_ t|_ t|_ dS)Nz+Offset and exposure are not yet implementedrr)NotImplementedErrorr!r"k_extra1k_extra2 _initializeHurdleCountResultsrHurdleCountResultsWrapperrL1HurdleCountResultsrL1HurdleCountResultsWrapperr) r)r#r$rdistzerodistrpzerorrr*msgr,s r-r"zHurdleCountModel.__init__Js  H$8?C%c** *              xE222.$=! 4(C%%%r.c|dvs|dvrtd|dkr't|j|jt|_n<|dkr6t|j|jt |_|xjdz c_|dkr!t|j|j|_ dS|dkr3t|j|j||_ |xj dz c_ dSdS)N)rnegbinz,dist and zerodist must be "poisson","negbin"r)rYr'r8r) rrr#r$r model1r rrmodel2rr)r)r"r#rr$s r-rzHurdleCountModel._initializecs - - - 555%'122 2 y $TZ'JJJDKK  ! !$TZ+<>>>DK MMQ MM 9  ,TZCCDKKK X  6tz499:<<.r.rqc|Srkrlrms r-roz&HurdleCountModel.fit..r0r. convergedrrr8cg|]}d|zS)zm_rl).0names r- z(HurdleCountModel.fit..sCCCD54<CCCr.) block_diagzneed covariancerl)rr(rr)rrrY mle_retvalsr1appendrrrrSrrr scipy.linalgr8normalized_cov_params cov_paramsstrrr)r)rrrsrtrurirvrxrrzr*results1results2rxnames1r8cov1cov2emodelfits r-rzHurdleCountModel.fits| { " "FGG G"4;?%7#kk #4;?%7#kk (## $+3+? +L+3+? +L+N;'!#8+<+C+3+<+C"E"E   H$5$>>   !2IqAAA  !2IqAAA   59 CCDK,BCCC$t{'== ,+++++04- $//11D$//11D4>JtT4J4JFO 1 1    A../....  $$T6?HhOO**844 s:AG G) G$$G)rc|}|dnd}||||\}}}d}| |r|j}n|}tt ||jz |jz dz |jz} |d| } || d} tj|| d|jj d|z|z} |j | |} |j j | | }d|z }tj| }|jj || }|dkr|tj| z|z S|d krtj| S|d kr| S|d krtj| |z S|d kr|S|d kr|S|dkrd|z S|dkrHtj| }|j| |\}}||z|d|z z|dzzz}|S|dkrL|j | |tj||d|}||dddfz}||dddf<|St#d|z)a Predict response variable or other statistic given exogenous variables. Parameters ---------- params : array_like The parameters of the model. exog : ndarray, optional Explanatory variables for the main count model. If ``exog`` is None, then the data from the model will be used. exog_infl : ndarray, optional Explanatory variables for the zero-inflation model. ``exog_infl`` has to be provided if ``exog`` was provided unless ``exog_infl`` in the model is only a constant. offset : ndarray, optional Offset is added to the linear predictor of the mean function with coefficient equal to 1. Default is zero if exog is not None, and the model offset if exog is None. exposure : ndarray, optional Log(exposure) is added to the linear predictor 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. Default is one if exog is is not None, and it is the model exposure if exog is None. which : str (optional) Statitistic to predict. Default is 'mean'. - 'mean' : the conditional expectation of endog E(y | x) - 'mean-main' : mean parameter of truncated count model. Note, this is not the mean of the truncated distribution. - 'linear' : the linear predictor of the truncated count model. - 'var' : returns the estimated variance of endog implied by the model. - 'prob-main' : probability of selecting the main model which is the probability of observing a nonzero count P(y > 0 | x). - 'prob-zero' : probability of observing a zero count. P(y=0 | x). This is equal to is ``1 - prob-main`` - 'prob-trunc' : probability of truncation of the truncated count model. This is the probability of observing a zero count implied by the truncation model. - 'mean-nonzero' : expected value conditional on having observation larger than zero, E(y | X, y>0) - 'prob' : probabilities of each count from 0 to max(endog), or for y_values if those are provided. This is a multivariate return (2-dim when predicting for several observations). y_values : array_like Values of the random variable endog at which pmf is evaluated. Only used if ``which="prob"`` Returns ------- predicted values Notes ----- 'prob-zero' / 'prob-trunc' is the ratio of probabilities of observing a zero count between hurdle model and the truncated count model. If this ratio is larger than one, then the hurdle model has an inflated number of zeros compared to the count model. If it is smaller than one, then the number of zeros is deflated. NTFrrr8)r$rrrz mean-nonzeroz prob-zeroz prob-mainz prob-truncrrr:rzwhich = %s is not available)lowerrr$r+r,rrr1rrr(r?r>rrTr)rr)r)rr$rrr;r<no_exog exog_zerok_zeros params_zero params_mainlin_predmu1rrmu2 prob_ntruncrmtvtr probs_mains r-r?zHurdleCountModel.predictsB ,$$E!%!9!9":""fh    ! I  s6{{T]2T]BaG-(XgX& WXX& F4-@dioa.@-@!ABB%&k!!+D!99K*88kJJ ] fXk,::3 LL F??rvh///+= = k ! !6(## # h  O n $ $6(##k1 1 k ! !  k ! !  l " "{? " e^^!!B[44["EEFBr>IY$?"a%$GGDK f__,,T26(#3#3V6!-##J )AAAtG, ,J(Jqqq!t  :UBCC Cr.)NrrrrNrrr)rrrrrrrr"rr5rrr?rrs@r-rrs0+ "  !# " # #1+#GZ,0*3(.DDDDDD2(1110=?,07;4444l #+CK2648~D~D~D~D~D~D~D~Dr.rceZdZedddzZdS)rz%A results class for Generic Truncatedone_line_description extra_attrN)rrrrrrlr.r-rrHs($ G((GGGr.rc4eZdZedddzZedZdS)rz%A results class for Truncated PoissonrUrVc|jjdkrd}t|tj|d}d|tj|dz z z S)Nr0dispersion is only available for zero-truncationrr;r8)rYr%rr1rTr?)r)r%rs r-_dispersion_factorz,TruncatedLFPoissonResults._dispersion_factorSs^ : q DC%c** * VDLLxL00 1 1B"&**q.))*r.Nrrrrrrr]rlr.r-rrNsI$ G((G++^+++r.rc4eZdZedddzZedZdS)rz/A results class for Truncated Negative BinomialrUrVc(|jjdkrd}t||jd}|jjj}t j|d}d|||dz zzt j||dz zdz z z S)Nrr[r=rr\r8) rYr%rrr>rr1rTr?)r)r%rrrs r-r]z3TruncatedNegativeBinomialResults._dispersion_factords : q DC%c** * B J ! 2 VDLLxL00 1 1EB1I%QqS ):):Q)>??@r.Nr^rlr.r-rr^sN$ =((G  A A^ A A Ar.rceZdZdS)rNrrrrlr.r-rrqDr.rceZdZdS)rNrbrlr.r-rrurcr.rceZdZdS)rNrbrlr.r-rr}rcr.rc^eZdZedddzZ d fd ZedZedZxZ S) rz A results class for Hurdle modelrUrVrbNct|||||||_||_|jjjdt|jz |_ dS)N)rxrrzr) r!r" results_zero results_countrYr#rr,rr) r)rYrrhrirxrrzr,s r-r"zHurdleCountResults.__init__sl       )* (.q1C 4D4DD r.cH|jjj|jjjzSrk)rhrllnullrir)s r-rkzHurdleCountResults.llnulls$!*1"+23 4r.cTtj|jj|jjSrk)r1r:rhbserirls r-rnzHurdleCountResults.bses y*.0B0FGGGr.)rbNN) rrrrrr"rrkrnrrs@r-rrs$ B((G =A E E E E E E44^4HH^HHHHHr.rceZdZdS)r Nrbrlr.r-r r rcr.r ceZdZdS)rNrbrlr.r-rrrcr.rceZdZdS)r!Nrbrlr.r-r!r!rcr.r!)8__all__r|numpyr1statsmodels.base.modelrrYstatsmodels.base.wrapperwrapperwrap#statsmodels.regression.linear_model regression linear_modellm"statsmodels.distributions.discreterr#statsmodels.discrete.discrete_modelrr r r r r rrstatsmodels.tools.numdiffrstatsmodels.tools.decoratorsrstatsmodels.tools.sm_exceptionsrcopyrrrrrrrr rrrrrrrRegressionResultsWrapperrpopulate_wrapperrrr rr!rlr.r-rs   %%%%%%%%%'''''''''000000000                    211111777777>>>>>>T<T<T<T<T<T<T<T+=         ;   /(***     ""=   1*,,,,,r.