!g@dZddlZddlmZddlmZmZGddZy)zL Created on Sun May 10 08:23:48 2015 Author: Josef Perktold License: BSD-3 N) NonePenalty)approx_fprime_cs approx_fprimeceZdZdZfdZd dZd fd Zd fd Zd dZd fd Z d fd Z d d Z d fd Z dfd Z xZS)PenalizedMixinaMixin class for Maximum Penalized Likelihood Parameters ---------- args and kwds for the model super class penal : None or instance of Penalized function class If penal is None, then NonePenalty is used. pen_weight : float or None factor for weighting the penalization term. If None, then pen_weight is set to nobs. TODO: missing **kwds or explicit keywords TODO: do we adjust the inherited docstrings? We would need templating to add the penalization parameters c|jdd|_|jdd|_t||i||jt |j |_|jt|_d|_|jjddgt|dg|_ |jjddgy)Npenal pen_weightr_null_drop_keys) popr r super__init__lenendogr _init_keysextendgetattrr )selfargskwds __class__s X/var/www/dash_apps/app1/venv/lib/python3.12/site-packages/statsmodels/base/_penalized.pyrzPenalizedMixin.__init__ sXXgt, 88L$7 $'$' ?? "!$**oDO :: $DJDO 67&t->C ##Wl$;<c n|2t|dr$|j|}|j|}|Sd}|S)N scaletyper)hasattrpredictestimate_scale)rparamsscalermus r _handle_scalezPenalizedMixin._handle_scale7sB =t[)\\&)++B/  rc | |j}t||fi|}|dk7r:|j|fi|}|d|z |z|jj |zz}|S)z3 Log-likelihood of model at params rr)r rlogliker#r func)rr r rllfr!rs rr%zPenalizedMixin.loglikeCso  Jgof-- ?&D&&v66E 1U7Z'$**//&*AA AC rc | |j}t||fi|}t|jd}|dk7r=|j |fi|}|d|z |z|z |j j|zz}|S)z@ Log-likelihood of model observations at params rr)r r loglikeobsfloatshaper#r r&)rr r rr'nobs_llfr!rs rr)zPenalizedMixin.loglikeobsQs  Jg 0401& ?&D&&v66E 1U7Z'(2TZZ__V5LL LC rc  jfd}|dk(r t||S|dk(rt||dStd)z4score based on finite difference derivative c.j|fdiSNr r%prr rs rz.PenalizedMixin.score_numdiff..gLDLLJzJTJrcsfdT)centeredz-method not recognized, should be "fd" or "cs")r rr ValueError)rr r methodrr%s` ` ` r score_numdiffzPenalizedMixin.score_numdiffasO  JJ T>#FG4 4 t^ 4@ @LM Mrc | |j}t||fi|}|dk7r:|j|fi|}|d|z |z|jj |zz}|S)z- Gradient of model at params rr)r rscorer#r deriv)rr r rscr!rs rr<zPenalizedMixin.scorepsq  J W]6 *T * ?&D&&v66E !E'J&)9)9&)AA AB rc | |j}t||fi|}t|jd}|dk7r=|j |fi|}|d|z |z|z |j j|zz}|S)z: Gradient of model observations at params rr)r r score_obsr*r+r#r r=)rr r rr>nobs_scr!rs rr@zPenalizedMixin.score_obs~s  J W v . . $ ?&D&&v66E !E'J&0DJJ4D4DV4LL LB rc P jfd}ddlm}|||S)z6hessian based on finite difference derivative c.j|fdiSr/r0r1s rr3z0PenalizedMixin.hessian_numdiff..r4rr) approx_hess)r statsmodels.tools.numdiffrD)rr r rr%rDs` `` rhessian_numdiffzPenalizedMixin.hessian_numdiffs)  JJ967++rc *| |j}t||fi|}|dk7rn|j|fi|}|jj |}|j dk(r#|d|z tj||zzz}|S|d|z |z|zz}|S)z, Hessian of model at params rr) r rhessianr#r deriv2ndimnpdiag)rr r rhessr!hrs rrHzPenalizedMixin.hessians  Jwv.. ?&D&&v66E !!&)Avv{%"''*q."999 %*,q00 rc ddlm}ddlm}t |||fr|j ddi|d}|d}t |d d|i|}|dur|S|durd }tjtj|j|kd}tjtj|j|kDd}|jr|j|fi|} | S|S) aminimize negative penalized log-likelihood Parameters ---------- method : None or str Method specifies the scipy optimizer as in nonlinear MLE models. trim : {bool, float} Default is False or None, which uses no trimming. If trim is True or a float, then small parameters are set to zero. If True, then a default threshold is used. If trim is a float, then it will be used as threshold. The default threshold is currently 1e-4, but it will change in future and become penalty function dependent. kwds : extra keyword arguments This keyword arguments are treated in the same way as in the fit method of the underlying model class. Specifically, additional optimizer keywords and cov_type related keywords can be added. r)GLMGam)GLMmax_start_irlsbfgsFr9Tg-C6?)*statsmodels.gam.generalized_additive_modelrP+statsmodels.genmod.generalized_linear_modelrQ isinstanceupdaterfitrKnonzeroabsr any _fit_zeros) rr9trimrrPrQres drop_index keep_indexres_auxrs rrYzPenalizedMixin.fits2 FC dS&M * KK)1- . >F <Dgk0040 5=Jt|BFF3::$6$=>qAJBFF3::$6$=>qAJ~~)$//*== r)N)Nr6)NN)__name__ __module__ __qualname____doc__rr#r%r)r:r<r@rFrHrY __classcell__)rs@rrr s@$=.   N  ,$99rr) rfnumpyrK _penaltiesrrErrrrTrrrjs!#EUUr