J/PhY8gdZddlZddlmZddlmZmZmZm Z m Z GddZ dZ dZ d Zd Zd Zd Zd ZdZGddZdZGddeZdZGddeZdZGddeZdZGddeZdZddZdS) )ContrastMatrix TreatmentPolySumHelmertDiffN) PatsyError)repr_pretty_delegaterepr_pretty_implsafe_issubdtype no_picklingassert_no_picklingc&eZdZdZdZeZdZeZ dS)raA simple container for a matrix used for coding categorical factors. Attributes: .. attribute:: matrix A 2d ndarray, where each column corresponds to one column of the resulting design matrix, and each row contains the entries for a single categorical variable level. Usually n-by-n for a full rank coding or n-by-(n-1) for a reduced rank coding, though other options are possible. .. attribute:: column_suffixes A list of strings to be appended to the factor name, to produce the final column names. E.g. for treatment coding the entries will look like ``"[T.level1]"``. ctj||_||_|jjdt |krt ddS)Nz(matrix and column_suffixes don't conform)npasarraymatrixcolumn_suffixesshapelenr )selfrrs O/var/www/html/test/jupyter/venv/lib/python3.11/site-packages/patsy/contrasts.py__init__zContrastMatrix.__init__*sOj(( . ; Q 3#7#7 7 7GHH H 8 7c@t|||j|jgdSN)r rr)rpcycles r _repr_pretty_zContrastMatrix._repr_pretty_2s$D4;0D"EFFFFFrN) __name__ __module__ __qualname____doc__rr __repr__r r __getstate__rrrrsJ&III $HGGGLLLrrcDtddgddggddg}tj|jtjdsJ|jddgksJt |ddl}|ttdgdggddgt|dS)Nrrab) rr array_equalreyerreprpytestraisesr r)cmr/s rtest_ContrastMatrixr28s !Q!Q(3* 5 5B >")RVAYY / //// #s + + + +HHHMMM MM*nsQCj3*EEErrct|tr|St|tr5 |dS#t$rt |cYSwxYwt |S)Nutf-8) isinstancestrbytesdecodeUnicodeDecodeErrorr.)objs r_obj_to_readable_strr;Ks}#s C   ::g&& &!   99    CyysAAAcd}|dd|dd|dd|dd|ddd|dd d dS) Ncdt|}t|tusJ||ksJdSr)r;typer6)r:expectedgots rtz$test__obj_to_readable_str..tXs7"3''CyyChrr1g?z1.0asdfu€r4z iso-8859-15zb'\xa4')encode)rAs rtest__obj_to_readable_strrEWs AaIIIAc5MMMAffAffAhoog)))Ahoom$$j11111rc fd|DS)Nc:g|]}dt|dS)[])r;).0levelprefixs r z _name_levels..ks1 Q Q Q 4U ; ; ; ; < Q Q Qrr')rLlevelss` r _name_levelsrOjs Q Q Q Q& Q Q QQrc:tdddgddgksJdS)Nr)r*cz[ab]z[ac])rOr'rrtest__name_levelsrRns, c3Z ( (VV,< < < < < < !II"64>::IfS[[1_%%I )QQQ 1c&kkAo*>!?!?YZZQRQRQR]AS T  T6*9*#5y1}8O#OPPi///rr) r!r"r#r$rrgrmr r&r'rrrrsQ<####### 0 0 0LLLrrct}|gd}|jgdksJtj|jgdgdgdgsJ|gd}|jddgksJtj|jddgd dgdd ggsJtd gd}|jd dgksJtj|jd dgddgdd ggsJtd gd}|jd dgksJtj|jd dgddgdd ggsJtd gd}|jd dgksJtj|jd dgddgdd ggsJtgd}|jddgksJtj|jddgd dgdd ggsJdS)Nr\)z[a]z[b]z[c])rrr)rrr)rrrz[T.b]z[T.c]rrrcz[T.a]r*r^z[T.1]z[T.0])rrgrrallcloserrmt1rs rtest_Treatmentrssh B  # #OOO 4 4F  !%:%:%: : : : : ;v}yyy)))YYY&G H HHHH  & & 7 7F  !gw%7 7 7 7 7 ;v}1v1v1v&> ? ????  # # # : :??? K KF  !gw%7 7 7 7 7 ;v}1v1v1v&> ? ????  $ $ $ ; ;OOO L LF  !gw%7 7 7 7 7 ;v}1v1v1v&> ? ????  % % % < <___ M MF  !gw%7 7 7 7 7 ;v}1v1v1v&> ? ????[[ / / : :F  !gw%7 7 7 7 7 ;v}1v1v1v&> ? ??????rc0eZdZdZddZdZdZdZeZ dS)raOrthogonal polynomial contrast coding. This coding scheme treats the levels as ordered samples from an underlying continuous scale, whose effect takes an unknown functional form which is `Taylor-decomposed`__ into the sum of a linear, quadratic, etc. components. .. __: https://en.wikipedia.org/wiki/Taylor_series For reduced-rank coding, you get a linear column, a quadratic column, etc., up to the number of levels provided. For full-rank coding, the same scheme is used, except that the zero-order constant polynomial is also included. I.e., you get an intercept column included as part of your categorical term. By default the levels are treated as equally spaced, but you can override this by providing a value for the `scores` argument. Examples: .. ipython:: python # Reduced rank dmatrix("C(a, Poly)", balanced(a=4)) # Full rank dmatrix("0 + C(a, Poly)", balanced(a=3)) # Explicit scores dmatrix("C(a, Poly([1, 2, 10]))", balanced(a=3)) This is equivalent to R's ``contr.poly``. (But note that in R, reduced rank encodings are always dummy-coded, regardless of what contrast you have set.) Nc||_dSrscores)rrws rrz Poly.__init__s  rcjt|}|j}|tj|}tj|t }t||kr#t d|dt|d||z}|dtj|dz}tj |\}}|tj tj |z}|tj tj|dzd z}d|ddd f<gd }|d td |Dz }|d|}|rt!||St!|ddddf|ddS)N)dtypeznumber of levels (z#) does not match number of scores ())r]r)rr]r+r)axisr) .Constant.Linear .Quadratic.Cubiccg|]}d|S)^r')rJis rrMz%Poly._code_either..s4441111,444r)rrwrarangerfloatr meanreshapelinalgqrsigndiagsqrtsumranger) r interceptrNnrwraw_polyqrrls r _code_eitherzPoly._code_eithers KK >Yq\\FF%000 v;;!  *,-AAs6{{{{<  &++-->>'**bill.B.B7.K.KKy||H%%1 RWRWQZZ   RWRVAqDq))) * **!!!Q$@@@ 44a 4444bqb   7!!U++ +"!AAAqrrE(E!""I66 6rc.|d|S)NTrrfs rrgzPoly.code_with_intercept s  v...rc.|d|S)NFrrfs rrmzPoly.code_without_intercept#s  ///rr) r!r"r#r$rrrgrmr r&r'rrrrsa  D777>///000LLLrrcHt}|gd}|jgdksJgdgdgdg}t|jt j|j|sJ|gd}|jddgksJt|jt j|jdd gd d gd d ggsJtgd gd}|jgdksJt|jt j|jgdgdgdgsJtgdgd}|jgdksJt|jt j|jgdgdgdgsJd dl}| ttd dgjgd|ttd}|jgdksJdS)Nr\)r|r}r~)r;f>, p ?)rr>, p )r;f?rr}r~rrrrr)r rv)rg gQ}.?)rg?Q?g3)rgGUg?guCz?)rr )rg9UgHA?)rg)U?gM lOb)rgWl?gQ?r)r|r}r~rz^4z^5) rrgrprintrrrprmr/r0r listr)rrrr?r/s r test_Polyr)s B  # #OOO 4 4F  !%K%K%K K K K K 655"""444H  &- ;v}h / ////  & & 7 7F  !i%> > > > > &- ; "$5 6 " # !#4 5    % % % 9 9/// J JF  !%K%K%K K K K K &- ; 7 7 7 7 7 7 6 6 6    % % % 9 9/// J JF  !%K%K%K K K K K &- ; 6 6 6 6 6 6 5 5 5   MMM MM*d1a&111EWWW  # #DqNN 3 3F  !&&&      rc6eZdZdZddZdZdZdZdZe Z dS) raDeviation coding (also known as sum-to-zero coding). Compares the mean of each level to the mean-of-means. (In a balanced design, compares the mean of each level to the overall mean.) For full-rank coding, a standard intercept term is added. One level must be omitted to avoid redundancy; by default this is the last level, but this can be adjusted via the `omit` argument. .. warning:: There are multiple definitions of 'deviation coding' in use. Make sure this is the one you expect before trying to interpret your results! Examples: .. ipython:: python # Reduced rank dmatrix("C(a, Sum)", balanced(a=4)) # Full rank dmatrix("0 + C(a, Sum)", balanced(a=4)) # Omit a different level dmatrix("C(a, Sum(1))", balanced(a=3)) dmatrix("C(a, Sum('a1'))", balanced(a=3)) This is equivalent to R's `contr.sum`. Nc||_dSromit)rrs rrz Sum.__init__s  rc^|jt|dz St||jS)Nr)rrrZrfs r_omit_iz Sum._omit_is, 9 v;;? "fdi00 0rc"t|}||}tj|dz }tj||dz f}|d|ddf|d|ddf<d||ddf<||dddf||dzdddf<|SNrr])rrrr-empty)rrNromit_ir-outs r _sum_contrastzSum._sum_contrasts KKf%%fQUmmh1q5z""gvgqqqj/GVGQQQJFAAAI"677AAA:FQJLL!!!O rc||}tjtjt ||jf}dg|jz}t||S)N[mean])rmr column_stackonesrrrr)rrNcontrastrrs rrgzSum.code_with_interceptsZ..v66"'#f++"6"6!HII#*x'??fo666rc||}||}|d|||dzdz}t|td|S)NrzS.)rrrrO)rrNrrincluded_levelss rrmzSum.code_without_intercepts_##F++f%% &/F6A:<<,@@fl4&I&IJJJrr) r!r"r#r$rrrrgrmr r&r'rrrrmsr:111777 KKK LLLrrc^t}|gd}|jgdksJtj|jgdgdgdgsJ|gd}|jddgksJtj|jdd gd dgd d ggsJ|gd }|jd d gksJtj|jdd gd dgd d ggsJtd}|gd}|jgdksJtj|jgdgdgdgsJ|gd}|jddgksJtj|jdd gd d gd dggsJ|gd}|jddgksJtj|jd d gdd gd dggsJtd}|gd}|jgdksJtj|jgdgdgdgsJ|gd}|jddgksJtj|jd d gdd gd dggsJtd}|gd}|jgdksJtj|jgdgdgdgsJ|gd}|jddgksJtj|jd d gdd gd dggsJdS)Nr\)r[S.a][S.b])rrr)rrrrr]r]rrrrr])r]ror_z[S.-1]z[S.-2]r)rr[S.c]r)rrr+z[S.0]z[S.2]r_)rrrr))rrgrrrprrm)rrrt2t3t4s rtest_Sumrs B  # #OOO 4 4F  !%A%A%A A A A A ;v}yyy)))[[[&I J JJJJ  & & 7 7F  !gw%7 7 7 7 7 ;v}1v1vBx&@ A AAAA  & &||| 4 4F  !h%9 9 9 9 9 ;v}1v1vBx&@ A AAAA !B  # #OOO 4 4F  !%A%A%A A A A A ;v}yyy+++yyy&I J JJJJ  & & 7 7F  !gw%7 7 7 7 7 ;v}1vBx!Q&@ A AAAA  & &yyy 1 1F  !gw%7 7 7 7 7 ;v}Bx!Q!Q&@ A AAAA "B  # #OOO 4 4F  !%A%A%A A A A A ;v}{{{IIIyyy&I J JJJJ  & & 7 7F  !gw%7 7 7 7 7 ;v}Bx!Q!Q&@ A AAAA #B  # #OOO 4 4F  !%A%A%A A A A A ;v}{{{IIIyyy&I J JJJJ  & & 7 7F  !gw%7 7 7 7 7 ;v}Bx!Q!Q&@ A AAAAAArc(eZdZdZdZdZdZeZdS)raNHelmert contrasts. Compares the second level with the first, the third with the average of the first two, and so on. For full-rank coding, a standard intercept term is added. .. warning:: There are multiple definitions of 'Helmert coding' in use. Make sure this is the one you expect before trying to interpret your results! Examples: .. ipython:: python # Reduced rank dmatrix("C(a, Helmert)", balanced(a=4)) # Full rank dmatrix("0 + C(a, Helmert)", balanced(a=4)) This is equivalent to R's `contr.helmert`. ct|}tj||dz f}tjd||ddtj|dz <d|tj|dz <|Sr)rrrjr diag_indices triu_indices)rrNrcontrs r_helmert_contrastzHelmert._helmert_contrastsm KK !QU$$,.IaOOabb "/!a%(()(*boa!e$$% rc tjtjt|||f}t ddgt |ddz}t||S)NH.rr)rrrrrrOrr)rrNrrs rrgzHelmert.code_with_interceptsq? WS[[ ! !4#9#9&#A#A B  'tk]T&*=M=M-MNNh888rc x||}t|td|ddS)Nrr)rrrOrrNrs rrmzHelmert.code_without_intercept s7))&11h T6!"":(F(FGGGrN) r!r"r#r$rrgrmr r&r'rrrrsO.,999HHHLLLrrc dt}gddfD]}||}|jgdksJtj|jgdgdgdgdgsJ||}|jgdksJtj|jgdgd gd gd gsJdS) Nr)r*rQd)z [H.intercept][H.b][H.c][H.d])rr]r]r])rrr]r])rrr+r])rrr)rrr)r]r]r]r)rr+r])rrr)rrgrrrprrm)rrrNrs r test_Helmertrs B''')=>   ''//%)U)U)UUUUU{ M __nnnmmm\\\ J     **622%)D)D)DDDDD{ MLLL+++zzz999M         rc(eZdZdZdZdZdZeZdS)raBackward difference coding. This coding scheme is useful for ordered factors, and compares the mean of each level with the preceding level. So you get the second level minus the first, the third level minus the second, etc. For full-rank coding, a standard intercept term is added (which gives the mean value for the first level). Examples: .. ipython:: python # Reduced rank dmatrix("C(a, Diff)", balanced(a=3)) # Full rank dmatrix("0 + C(a, Diff)", balanced(a=3)) c6t|}tj||dz f}tjd|}tj||}tj|dz \}}tj|}||z t|z |||||f<tj||ddd} tj|dz \}}tj|}| t|z |||dz||f<|Sr) rrrjrrepeatrargsortr tril_indices) rrNnlevelsr int_range upper_introw_icol_i col_order lower_ints r_diff_contrastzDiff._diff_contrast4sf++'7Q;/00Ia)) Ii33 w{33 uJu%% 5>5HE M M 5 eIi 001Ii44R499 w{33 uJu%% 8AE'NN8ReI"E)$445 rctjtjt|||f}t |t d|S)ND.)rrrrrrrOrs rrgzDiff.code_with_interceptGsN?BGCKK$8$8$:M:Mf:U:U#VWWh T6(B(BCCCrc x||}t|td|ddS)Nrr])rrrOrs rrmzDiff.code_without_interceptKs7&&v..h T6#2#;(G(GHHHrN) r!r"r#r$rrgrmr r&r'rrrr sR&&DDDIIILLLrrcZt}|gd}|jgdksJtj|jgdgdgdgdgsJ|gd}|jgdksJtj|jgdgd gd gd gsJdS) Nr)[D.a][D.b][D.c]z[D.d])rп)r?rr)rr?r)rrr?)rrr)rrr)rrr)rrr)rrr)rrgrrrprrmrqs r test_diffrRs B  # #$8$8$8 9 9F  !%I%I%I I I I I ; - - - , , , + + + * * *     & &';';'; < !3D-E-EFF   7++F333..v666rr) __all__numpyrpatsyr patsy.utilr r r r rrr2r;rErOrRrUrZr`rrsobjectrrrrrrrrrr'rrrss L K KD   &   222&RRR===III D D D ; ; ;11111111h@@@0KKKKK6KKK\AAAH>>>>>&>>>B#B#B#BL99999f999x    /////6///dB777777r