M/PhP*ddlZddlmZ ddZdZdS)N)PHRegTc tj|}||}||}||ddf}t|}tj|d\}} tj||ddz } t |} |dktj} ||| z }g} dg|z}dg|z}dg|z}t| D]}||dzktj}|rt||| }| j}t||d|z  }| j}tj|dddf|dddff}||dz}||dz}n|}d}t|D]}|dkr|||ddfz }||}|||<|dkrtj||dddddd}||| z}||z }d }tj||k||kz}d||<d|z }tj|d tj}tj|} tj| | }!tj|!}"tjd|"ddf||<|| ||<tj| |||zt| }#tj|||#z||z }$t|$tj||ddddz }%|%t|$kr|$|%dz |$|%d<|||$z }||||$zz }|||z}| ||| fS) aI Calculates cumulative incidence functions using kernels. Parameters ---------- time : array_like The observed time values status : array_like The status values. status == 0 indicates censoring, status == 1, 2, ... are the events. exog : array_like Covariates such that censoring becomes independent of outcome times conditioned on the covariate values. kfunc : function A kernel function freq_weights : array_like Optional frequency weights dimred : bool If True, proportional hazards regression models are used to reduce exog to two columns by predicting overall events and censoring in two separate models. If False, exog is used directly for calculating kernel weights without dimension reduction. NT)return_inverserightsiderV瞯>B3wwtUBF33vd||yr*VD\\aCRCj)1!"Iq UGBqEM'*5zz333B )BqEBJ233CSBOF1IdddOQ???BCHH}}rAv;BCC#s |A,,   4KC # "9ct|||}|j}t||d|z }|j}t j|dddf|dddff} t |} t j|dk} t j|| } t j |} || }|| }| | ddf} t j || ddz }||| z }d}t| D]}| | |ddfz }||}t j |dddddd}||z}||z }d}t j||k||kz} d|| <d|z }t j|d tj}t j|}t j ||}t j|}|||z }||||zz }||| z}|| fS) a Estimate the marginal survival function under dependent censoring. Parameters ---------- time : array_like The observed times for each subject status : array_like The status for each subject (1 indicates event, 0 indicates censoring) exog : array_like Covariates such that censoring is independent conditional on exog kfunc : function Kernel function freq_weights : array_like Optional frequency weights Returns ------- probs : array_like The estimated survival probabilities times : array_like The times at which the survival probabilities are estimated References ---------- Zeng, Donglin 2004. Estimating Marginal Survival Function by Adjusting for Dependent Censoring Using Many Covariates. The Annals of Statistics 32 (4): 1533 55. doi:10.1214/009053604000000508. https://arxiv.org/pdf/math/0409180.pdf r Nrrgr r rg-q=)rrrrrrrr rrrrrrr!r"r#r$)r(r)r*r+r,r;r<r=r>r?nixdr0r.r2sprobrAr8rCrDrErFrGrHrIs rN_kernel_survfuncrUwsWH dF # # ' ' ) )C{{}}-H dAJ ' ' + + - -C{{}}-H YD)8AAAtG+<= > >F D A .1 % %C Id3i E D  B 8D BZF BE]F u7 3 3 3a 7B#l&6&6&8&88 E 1XX,, fQTl " U2YY "TTrT(##DDbD)6kEk  ^URZC"H5 6 6B3wwtUBF++vd||yr"vd||   TMEE TLO+ +EE   %<rP)T)numpyr&statsmodels.duration.hazard_regressionrrOrUrPrNrYs_888888!%oooodWWWWWrP