o þ0Gfmã@stdZddlZddlmZddlmZddlmZdd„Z dd „Z Gd d „d eƒZ Gd d „d eƒZ Gdd„deƒZ dS)aµ Created on Mon Jul 26 08:34:59 2010 Author: josef-pktd changes: added offset and zero-inflated version of Poisson - kind of ok, need better test cases, - a nan in ZIP bse, need to check hessian calculations - found error in ZIP loglike - all tests pass with Issues ------ * If true model is not zero-inflated then numerical Hessian for ZIP has zeros for the inflation probability and is not invertible. -> hessian inverts and bse look ok if row and column are dropped, pinv also works * GenericMLE: still get somewhere (where?) "CacheWriteWarning: The attribute 'bse' cannot be overwritten" * bfgs is too fragile, does not come back * `nm` is slow but seems to work * need good start_params and their use in genericmle needs to be checked for consistency, set as attribute or method (called as attribute) * numerical hessian needs better scaling * check taking parts out of the loop, e.g. factorial(endog) could be precalculated éN)Ústats)Ú factorial)ÚGenericLikelihoodModelcCst t ||¡¡S)N©ÚnpÚmaxÚabs©Zarr1Zarr2©r ú|jt |j|¡}|j}t |¡||t t|ƒ¡}|Sr)r+rrrrrrr)rrrrÚnloglikr r r r|s zPoissonOffsetGMLE.nloglikeobs©NNr&©r r!r"r#r-rÚ __classcell__r r r0r r%_s r%cr$) Ú PoissonZiGMLEaÊMaximum Likelihood Estimation of Poisson Model This is an example for generic MLE which has the same statistical model as discretemod.Poisson but adds offset and zero-inflation. Except for defining the negative log-likelihood method, all methods and results are generic. Gradients and Hessian and all resulting statistics are based on numerical differentiation. There are numerical problems if there is no zero-inflation. Nr&c s²d|_tƒj||f|dgdœ|¤Ž|dur)|jdkr#|dd…df}| ¡|_nd|_|dur9t |jdf¡|_ |j j d|_ t  t |j ¡df¡|_ |j d7_ dg|_dS)Nr Zzi)r(Zextra_params_namesr'rÚ start_params)Zk_extrar,r-r)r*r+rZonesZnobsrÚshapeZnparamsZhstackr7Z cloneattrr.r0r r r-¢s"ÿÿ   zPoissonZiGMLE.__init__cCs˜|dd…}ddt |d¡}|jt |j|¡}|j}t d|¡ t |¡||t t|ƒ¡}t |t ||dk ¡¡ ||dk<|S)rNéÿÿÿÿr r)rrr+rrrrr)rrZbetaZgammrrr2r r r r½s 0(zPoissonZiGMLE.nloglikeobsr3r4r r r0r r6“sr6)r#ZnumpyrZscipyrZ scipy.specialrZstatsmodels.base.modelrr rrr%r6r r r r Ús   34