U md@sdZddlZddlmZddlmZddlm Z GdddZ e dkrdgZ de krd Z ejejje d fd ee dffZejeed d ded ddd fjZeddddgddddgddddggZeddddgddddgddddggZeeeZedejjejd 7ZeedddgZedejjejd Ze eeZee edej!ddddee dS)zQ Created on Sun Nov 14 08:21:41 2010 Author: josef-pktd License: BSD (3-clause) N)pca) LeaveOneOutc@s<eZdZdZddZdddZd d Zdd d Zd dZdS)FactorModelUnivariatea  Todo: check treatment of const, make it optional ? add hasconst (0 or 1), needed when selecting nfact+hasconst options are arguments in calc_factors, should be more public instead cross-validation is slow for large number of observations cCst||_t||_dS)N)npasarrayendogexog)selfrrr b/home/sam/Atlas/atlas_env/lib/python3.8/site-packages/statsmodels/sandbox/datarich/factormodels.py__init__s zFactorModelUnivariate.__init__NrTcCsn|dkr|j}n t|}t||dd\}}}}||_|rRtj|dd|_d|_n ||_d|_||_ ||_ dS)zget factor decomposition of exogenous variables This uses principal component analysis to obtain the factors. The number of factors kept is the maximum that will be considered in the regression. N)keepdim normalizeT)prependr) rrrrZ exog_reducedsmZ add_constantfactorshasconstevalsevecs)r xrZaddconstZxredfactrrr r r calc_factors!s z"FactorModelUnivariate.calc_factorscCs:t|ds|t|j|jddd|dfS)NZfactors_wconstr )hasattrrrOLSrrfit)r Znfactr r r fit_fixed_nfact8s z%FactorModelUnivariate.fit_fixed_nfactc Cst|ds||j}|dkr0|jjd|}||dkrDtdt|d}|j}g}td||D]}|jddd|f}t || } |s |dkrt t |}d} |D]T\} } t || || ddf } | || | j| j|| ddfd7} qntj} ||| j| j| j| gqft||_}tjt|ddddfd t|dddfd t|ddd fd f|_dS) aWestimate the model and selection criteria for up to maxfact factors The selection criteria that are calculated are AIC, BIC, and R2_adj. and additionally cross-validation prediction error sum of squares if `skip_crossval` is false. Cross-validation is not used by default because it can be time consuming to calculate. By default the cross-validation method is Leave-one-out on the full dataset. A different cross-validation sample can be specified as an argument to cv_iter. Results are attached in `results_find_nfact` rNr zFnothing to do, number of factors (incl. constant) should be at least 1 @r)rrrrshape ValueErrorminrrangerrrrlenmodelZpredictparamsrnanappendZaicZbicZ rsquared_adjarrayresults_find_nfactZr_ZargminZargmax best_nfact)r maxfact skip_crossvalcv_iterry0resultskrresZprederr2ZinidxZoutidxZres_l1or r r fit_find_nfact=s<        4z$FactorModelUnivariate.fit_find_nfactcCst|ds||j}d}|d7}|ddt|j7}ddlm}dd }d gd gd }t|d }|||d|d}|d7}|d7}|d| 7}|d7}|d7}|d7}|d7}|S)zprovides a summary for the selection of the number of factors Returns ------- sumstr : str summary of the results for selecting the number of factors r,z, Best result for k, by AIC, BIC, R2_adj, L1Oz z%5d %4d %6d %5dr) SimpleTablezk, AIC, BIC, R2_adj, L1Oz, z%6dz%10.3f)Z data_fmtsN)Ztxt_fmtz" PCA regression on simulated data,z+ DGP: 2 factors and 4 explanatory variables z) Notes: k is number of components of PCA,z& constant is added additionallyz- k=0 means regression on constant onlyz? L1O: sum of squared prediction errors for leave-one-out) rr5r,tupler-Zstatsmodels.iolib.tabler7splitdict__str__)r r2Zsumstrr7headersZ numformatZtxt_fmt1Ztablr r r summary_find_nfacts&    z(FactorModelUnivariate.summary_find_nfact)NrT)NTN) __name__ __module__ __qualname____doc__r rrr5r?r r r r rs   Br__main__r i)sizer8r!g?rg@rg?g?zwith cross validation - slowerF)r.r/r0)"rCnumpyrZstatsmodels.apiapirZstatsmodels.sandbox.toolsrZ#statsmodels.sandbox.tools.cross_valrrr@ZexamplesZnobsZc_randomnormalZonesZf0repeateyeZarangeTZf2xcoefr+dotZx0r"Zytruer1modprintr?r5r r r r s:     &0