U Kv fý#ã @sÖdZddlmZddlZddlmZd?dd„Zdd„Z d d „Z d Z d Z d d„Z dd„Zdd„ZedkrÒdZejjdedfd defdefdefdefg¡Zejjedfd defdefg¡ZddlZejjjeefdd Ze ed!fe¡Z ej!j"D]Z#ee#e e#<qøej!j"D]Z#ee#e e#<qed"e ƒ\Z$Z"ed#e ƒ\Z$Z"e 'd$d%„e"Dƒ¡Z(e( )d!¡d&ejjedZ*e +e*e(¡ ,¡Z-e&e-j.ƒe&e e e-ƒƒe 'd'd%„ed(e"ƒDƒ¡Z(e( )d!¡d&ejjedZ*e +e*e(¡ ,¡Z-e&e-j.ƒe&e e e-ƒƒe !d)efd*efd+efd,efd-efd.efd/efd0efg¡Z/ej0d1e/d2dd3Z1e&d4d5d%„e1j!j"Dƒƒe1j2 e3¡Z4e4 5d6e6e1j!j"ƒ¡ 7d!¡Z8e1e8dd…fj9Z:e&e:j;ƒe&e:j!ƒed7e:ƒ\Ze:d0Z?e +e?e>¡ ,¡Z@e&e@j.ƒe&e e e@ƒƒd: Ae1j!j"dd6…¡ZBed;e:ƒ\ZCZDe 'dd%„eeHeDƒDƒ¡ZIe:d0ZJe +eJeI¡ ,¡ZKe&e e eKƒƒq|dS)@a convenience functions for ANOVA type analysis with OLS Note: statistical results of ANOVA are not checked, OLS is checked but not whether the reported results are the ones used in ANOVA includes form2design for creating dummy variables TODO: * ... * é)ÚlmapNFcCs^| ¡}t |¡}|r0|dd…df|k t¡S|dd…df|k t¡dd…dd…fSdS)z|convert array of categories to dummy variables by default drops dummy variable for last category uses ravel, 1d onlyNéÿÿÿÿ)ZravelÚnpÚuniqueÚastypeÚint)ÚxÚ returnallÚgroups©r úP/tmp/pip-unpacked-wheel-2v6byqio/statsmodels/sandbox/regression/try_ols_anova.pyÚ data2dummys  r cCsLt tt| ¡ƒ¡}||dd…ddd…fk d¡j t¡dd…dd…fS)zÝcreates product dummy variables from 2 columns of 2d array drops last dummy variable, but not from each category singular with simple dummy variable but not with constant quickly written, no safeguards Nr) rrrÚtupleÚtolistÚallÚTrr)rr r r r Údata2proddummys rcCs.|jdkr|dd…df}t|dd}||S)zýcreate dummy continuous variable Parameters ---------- x1 : 1d array label or group array x2 : 1d array (float) continuous variable Notes ----- useful for group specific slope coefficients in regression éNT)r )Úndimr )Úx1Zx2Údummyr r r Údata2groupcont.s  raW ANOVA statistics (model sum of squares excludes constant) Source DF Sum Squares Mean Square F Value Pr > F Model %(df_model)i %(ess)f %(mse_model)f %(fvalue)f %(f_pvalue)f Error %(df_resid)i %(ssr)f %(mse_resid)f CTotal %(nobs)i %(uncentered_tss)f %(mse_total)f R squared %(rsquared)f a] ANOVA statistics (model sum of squares includes constant) Source DF Sum Squares Mean Square F Value Pr > F Model %(df_model)i %(ssmwithmean)f %(mse_model)f %(fvalue)f %(f_pvalue)f Error %(df_resid)i %(ssr)f %(mse_resid)f CTotal %(nobs)i %(uncentered_tss)f %(mse_total)f R squared %(rsquared)f c Csbi}| |j¡ddddddddd d d g }|D]}t||ƒ||<q.|jj|d <|j|j|d <|S)zjupdate regression results dictionary with ANOVA specific statistics not checked for completeness Zdf_modelZdf_residZessÚssrÚuncentered_tssZ mse_modelZ mse_residZ mse_totalZfvalueZf_pvalueZrsquaredÚnobsZ ssmwithmean)ÚupdateÚ__dict__ÚgetattrÚmodelrrr)ÚresÚadZ anova_attrÚkeyr r r Ú anovadict[s  þ r"cCshi}g}| ¡D]L}|dkr>t |jd¡|d<| d¡qd|kr^||||<| |¡q|dd…dkr˜| d¡d}t||ƒ||<| |¡q|dd…d krú| d¡d d ¡}ttj||d||dfƒ|d  |¡<| d  |¡¡q|dd…d krV| d¡d d ¡}t ||d||dƒ|d  |¡<| d  |¡¡qt d ƒ‚q||fS)aconvert string formula to data dictionary ss : str * I : add constant * varname : for simple varnames data is used as is * F:varname : create dummy variables for factor varname * P:varname1*varname2 : create product dummy variables for varnames * G:varname1*varname2 : create product between factor and continuous variable data : dict or structured array data set, access of variables by name as in dictionaries Returns ------- vars : dictionary dictionary of variables with converted dummy variables names : list list of names, product (P:) and grouped continuous variables (G:) have name by joining individual names sorted according to input Examples -------- >>> xx, n = form2design('I a F:b P:c*d G:c*f', testdata) >>> xx.keys() ['a', 'b', 'const', 'cf', 'cd'] >>> n ['const', 'a', 'b', 'cd', 'cf'] Notes ----- with sorted dict, separate name list would not be necessary ÚIrÚconstú:NézF:rzP:Ú*ÚzG:zunknown expression in formula) ÚsplitrZonesÚshapeÚappendr rZc_ÚjoinrÚ ValueError)ÚssÚdataÚvarsÚnamesÚitemÚvr r r Ú form2designls.$    ,$ r4cCs(|dd…}| ¡D]}| |¡q|S)zwdrop names from a list of strings, names to drop are in space delimited list does not change original list N)r)Úremove)r.ÚliZnewlir2r r r Údropname©s   r7Ú__main__ièéé)ÚsizeÚaÚbÚcÚdr&ÚeÚfT)Úflattenrz I a F:b P:c*dzI a F:b P:c*d G:a*e fcCsg|] }t|‘qSr ©Úxx©Ú.0Únnr r r Ú ÔsrHg{®Gáz„?cCsg|] }t|‘qSr rCrEr r r rHÜszae fZbreedZsexZlitterZpenZpigZageZbageÚyz dftest3.dataÚ.)ÚmissingZusemaskrKcCsg|]}tj| ¡‘qSr )ÚdtaÚmaskÚsum)rFÚkr r r rHñsrz I F:sex agecCsg|] }t|‘qSr )Úxx_b1rEr r r rHsr(ú z'I F:breed F:sex F:litter F:pen age bagecCsg|] }t|‘qSr ©Úxx_b1arEr r r rHsz Results droppingcCsg|] }t|‘qSr rRrEr r r rHs)F)LÚ__doc__Zstatsmodels.compat.pythonrZnumpyrZstatsmodels.apiÚapiÚsmr rrZ anova_str0Z anova_strr"r4r7Ú__name__rÚrandomÚrandintÚviewrZ testdataintÚnormalÚfloatZ testdatacontZnumpy.lib.recfunctionsÚlibZ recfunctionsZ zip_descrZdt2ÚemptyZtestdataZdtyper1ÚnamerDÚnÚprintZ column_stackÚXrNrIZOLSZfitZrest1ÚparamsZdt_bZ genfromtxtrLrMÚboolÚmZreshapeÚlenÚanyZdroprowsr/Z dta_use_b1r*rPZnames_b1ZX_b1Zy_b1Zrest_b1r,ZallexogrSZ names_b1aZX_b1aZy_b1aZrest_b1aZdropnZX_b1a_Zy_b1a_Z rest_b1a_r r r r Ús†     = 2$     þ