/DjYgdZddlZddlmZddlmZmZmZm Z m Z GddZ dZ dZ d Zd Zd Zd Zd ZdZGddZdZGddeZdZGddeZdZGddeZdZGddeZdZddZy) )ContrastMatrix TreatmentPolySumHelmertDiffN) PatsyError)repr_pretty_delegaterepr_pretty_implsafe_issubdtype no_picklingassert_no_picklingc$eZdZdZdZeZdZeZ y)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 |k7r t dy)Nz(matrix and column_suffixes don't conform)npasarraymatrixcolumn_suffixesshapelenr )selfrrs SC:\Crop_Prediction\Backend\crop-ai-system\venv\Lib\site-packages\patsy/contrasts.py__init__zContrastMatrix.__init__*sGjj( . ;;  Q 3#7 7GH H 8cJt|||j|jgyN)r rr)rpcycles r _repr_pretty_zContrastMatrix._repr_pretty_2sD4;;0D0D"EFrN) __name__ __module__ __qualname____doc__rr __repr__r r __getstate__rrrrs!&I $HGLrrc4tddgddggddg}tj|jtjdsJ|j ddgk(sJt |ddl}|jttdgdggddgt|y)Nrrab) rr array_equalreyerreprpytestraisesr r)cmr/s rtest_ContrastMatrixr28s !Q!Q(3* 5B >>"))RVVAY //  #s ++H MM*nsQCj3*Errct|tr|St|tr |jdSt |S#t$rt |cYSwxYw)Nutf-8) isinstancestrbytesdecodeUnicodeDecodeErrorr.)objs r_obj_to_readable_strr;KsV#s C  ::g& &Cy" 9  sAAAcd}|dd|dd|dd|dd|djdd|djd d y) NcNt|}t|tusJ||k(sJyr)r;typer6)r:expectedgots rtz$test__obj_to_readable_str..tXs'"3'CyChrr1g?z1.0asdfu€r4z iso-8859-15zb'\xa4')encode)rAs rtest__obj_to_readable_strrEWsY aIc5Mffffhoog)hoom$j1rcJ|Dcgc]}d|t|dc}Scc}w)N[])r;)prefixlevelslevels r _name_levelsrLjs!JP Q 4U ; < QQ Qs c.tdddgddgk(sJy)Nr)r*cz[ab]z[ac])rLr'rrtest__name_levelsrOns c3Z (VV,< << > !I"64>>:IffS[1_%II )Q 1c&kAo*>!?YZQR]AS T T6*9#5y1}8O#OPi//rr) r!r"r#r$rrdrjr r&r'rrrrs<## 0Lrrc0t}|jgd}|jgdk(sJtj|j gdgdgdgsJ|j gd}|jddgk(sJtj|j ddgd dgdd ggsJtd j gd}|jd dgk(sJtj|j d dgddgdd ggsJtd j gd}|jd dgk(sJtj|j d dgddgdd ggsJtd j gd}|jd dgk(sJtj|j d dgddgdd ggsJtj gd}|jddgk(sJtj|j ddgd dgdd ggsJy)NrY)z[a]z[b]z[c])rrr)rrr)rrrz[T.b]z[T.c]rrr`z[T.a]r*r[z[T.1]z[T.0])rrdrrallcloserrjt1rs rtest_Treatmentrps B  # #O 4F  ! !%: :: ;;v}}y)Y&G HH  & & 7F  ! !gw%7 77 ;;v}}1v1v1v&> ??  # : :? KF  ! !gw%7 77 ;;v}}1v1v1v&> ??  $ ; ;O LF  ! !gw%7 77 ;;v}}1v1v1v&> ??  % < <_ MF  ! !gw%7 77 ;;v}}1v1v1v&> ??[ / / :F  ! !gw%7 77 ;;v}}1v1v1v&> ?? ?rc.eZdZdZddZdZdZdZeZ y)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||_yrscores)rrts rrz Poly.__init__s  rc,t|}|j}|tj|}tj|t }t||k7rt d|dt|d||jz}|jdtj|jdz}tjj|\}}|tjtj|z}|tjtj|dzd z}d|ddd f<gd }|td |D cgc]} d |  c} z }|d|}|r t!||St!|ddddf|ddScc} w)N)dtypeznumber of levels (z#) does not match number of scores ())rZr)rrZr+r)axisr) .Constant.Linear .Quadratic.Cubic^)rrtrarangerfloatr meanreshapelinalgqrsigndiagsqrtsumranger) r interceptrJnrtraw_polyqrriis r _code_eitherzPoly._code_eithersV K >YYq\FF%0 v;! ,-s6{<  &++->>'*biil.B.B7.KKyy||H%1 RWWRWWQZ   RWWRVVAqDq) **!Q$@ a 411,44bq  !!U+ +"!AqrE(E!"I6 65s Fc&|jd|S)NTrrcs rrdzPoly.code_with_intercept s  v..rc&|jd|S)NFrrcs rrjzPoly.code_without_intercept#s  //rr) r!r"r#r$rrrdrjr r&r'rrrrs# D7>/0LrrcNt}|jgd}|jgdk(sJgdgdgdg}t|jt j |j|sJ|jgd}|jddgk(sJt|jt j |jdd gd d gd d ggsJtgd jgd}|jgdk(sJt|jt j |jgdgdgdgsJtgdjgd}|jgdk(sJt|jt j |jgdgdgdgsJd dl}|jttd dgjgd|jttd}|jgdk(sJy)NrY)ryrzr{)r;f>, p ?)rr>, p )r;f?rrzr{rrrrr)r rs)rg gQ}.?)rg?Q?g3)rgGUg?guCz?)rr )rg9UgHA?)rg)U?gM lOb)rgWl?gQ?r)ryrzr{r|z^4z^5) rrdrprintrrrmrjr/r0r listr)rorr?r/s r test_Polyr)s B  # #O 4F  ! !%K KK 6"4H  &-- ;;v}}h //  & & 7F  ! !i%> >> &-- ;; "$5 6 " # !#4 5   % 9 9/ JF  ! !%K KK &-- ;; 7 7 6   % 9 9/ JF  ! !%K KK &-- ;; 6 6 5   MM*d1a&1EEW  # #DqN 3F  ! !&  rc4eZdZdZddZdZdZdZdZe Z y) raDeviation coding (also known as sum-to-zero coding). Compares the mean of each level to the mean-of-means. (In a balanced design, compares the mean of each level to the overall mean.) For full-rank coding, a standard intercept term is added. One level must be omitted to avoid redundancy; by default this is the last level, but this can be adjusted via the `omit` argument. .. warning:: There are multiple definitions of 'deviation coding' in use. Make sure this is the one you expect before trying to interpret your results! Examples: .. ipython:: python # Reduced rank dmatrix("C(a, Sum)", balanced(a=4)) # Full rank dmatrix("0 + C(a, Sum)", balanced(a=4)) # Omit a different level dmatrix("C(a, Sum(1))", balanced(a=3)) dmatrix("C(a, Sum('a1'))", balanced(a=3)) This is equivalent to R's `contr.sum`. Nc||_yromit)rrs rrz Sum.__init__s  rcb|jt|dz St||jS)Nr)rrrWrcs r_omit_iz Sum._omit_is+ 99 v;? "fdii0 0rc t|}|j|}tj|dz }tj||dz f}|d|ddf|d|ddf<d||ddf<||dddf||dzdddf<|SNrrZ)rrrr-empty)rrJromit_ir-outs r _sum_contrastzSum._sum_contrasts Kf%ffQUmhh1q5z"gvgqj/GVGQJFAI"67A:FQJL!O rc|j|}tjtjt ||j f}dg|j z}t||S)N[mean])rjr column_stackonesrrrr)rrJcontrastrrs rrdzSum.code_with_interceptsU..v6"''#f+"6!HI#*x'?'??fo66rc|j|}|j|}|d|||dzdz}t|td|S)NrzS.)rrrrL)rrJrrincluded_levelss rrjzSum.code_without_interceptsO##F+f% &/F6A:<,@@fl4&IJJrr) r!r"r#r$rrrrdrjr r&r'rrrrms(:17 K Lrrct}|jgd}|jgdk(sJtj|j gdgdgdgsJ|j gd}|jddgk(sJtj|j dd gd dgd d ggsJ|j gd }|jd d gk(sJtj|j dd gd dgd d ggsJtd}|jgd}|jgdk(sJtj|j gdgdgdgsJ|j gd}|jddgk(sJtj|j dd gd d gd dggsJ|j gd}|jddgk(sJtj|j d d gdd gd dggsJtd}|jgd}|jgdk(sJtj|j gdgdgdgsJ|j gd}|jddgk(sJtj|j d d gdd gd dggsJtd}|jgd}|jgdk(sJtj|j gdgdgdgsJ|j gd}|jddgk(sJtj|j d d gdd gd dggsJy)NrY)r[S.a][S.b])rrr)rrrrrZrZrrrrrZ)rZrlr\z[S.-1]z[S.-2]r)rr[S.c]r)rrr+z[S.0]z[S.2]r\)rrrr))rrdrrrmrrj)rort2t3t4s rtest_Sumrs B  # #O 4F  ! !%A AA ;;v}}y)[&I JJ  & & 7F  ! !gw%7 77 ;;v}}1v1vBx&@ AA  & &| 4F  ! !h%9 99 ;;v}}1v1vBx&@ AA !B  # #O 4F  ! !%A AA ;;v}}y+y&I JJ  & & 7F  ! !gw%7 77 ;;v}}1vBx!Q&@ AA  & &y 1F  ! !gw%7 77 ;;v}}Bx!Q!Q&@ AA "B  # #O 4F  ! !%A AA ;;v}}{Iy&I JJ  & & 7F  ! !gw%7 77 ;;v}}Bx!Q!Q&@ AA #B  # #O 4F  ! !%A AA ;;v}}{Iy&I JJ  & & 7F  ! !gw%7 77 ;;v}}Bx!Q!Q&@ AA Arc&eZdZdZdZdZdZeZy)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)rrrgr diag_indices triu_indices)rrJrcontrs r_helmert_contrastzHelmert._helmert_contrastsf K !QU$,.IIaOab "//!a%()(*booa!e$% rc tjtjt||j |f}t ddgt |ddz}t||S)NH.rr)rrrrrrLrr)rrJrrs rrdzHelmert.code_with_intercepts]?? WWS[ !4#9#9&#A B 'tk]T&*=M-MNh88rc V|j|}t|td|ddS)Nrr)rrrLrrJrs rrjzHelmert.code_without_intercept s,))&1h T6!":(FGGrN) r!r"r#r$rrdrjr r&r'rrrrs.,9HLrrc vt}gddfD]}|j|}|jgdk(sJtj|j gdgdgdgdgsJ|j |}|jgdk(sJtj|j gdgd gd gd grJy) Nr)r*rNd)z [H.intercept][H.b][H.c][H.d])rrZrZrZ)rrrZrZ)rrr+rZ)rrr)rrr)rZrZrZr)rr+rZ)rrr)rrdrrrmrrj)rorJrs r test_Helmertrs B')=>  ''/%%)UUU{{ MM nm\ J  **62%%)DDD{{ MML+z9M    rc&eZdZdZdZdZdZeZy)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)) ct|}tj||dz f}tjd|}tj||}tj |dz \}}tj |}||z t|z |||||f<tj||ddd} tj|dz \}}tj |}| t|z |||dz||f<|Sr) rrrgrrepeatrargsortr tril_indices) rrJnlevelsr int_range upper_introw_icol_i col_order lower_ints r_diff_contrastzDiff._diff_contrast4sf+'7Q;/0IIa) IIi3 w{3 uJJu% 5>5HE M 5 eIi 001IIi4R49 w{3 uJJu% 8AE'N8ReI"E)$445 rctjtjt||j |f}t |t d|S)ND.)rrrrrrrLrs rrdzDiff.code_with_interceptGsA??BGGCK$8$:M:Mf:U#VWh T6(BCCrc V|j|}t|td|ddS)NrrZ)rrrLrs rrjzDiff.code_without_interceptKs,&&v.h T6#2;(GHHrN) r!r"r#r$rrdrjr r&r'rrrr s&&DILrrcht}|jgd}|jgdk(sJtj|j gdgdgdgdgsJ|j gd}|jgdk(sJtj|j gdgd gd gd gsJy) Nr)[D.a][D.b][D.c]z[D.d])rп)r?rr)rr?r)rrr?)rrr)rrr)rrr)rrr)rrr)rrdrrrmrrjrns r test_diffrRs B  # #$8 9F  ! !%I II ;; - , + *    & &'; >!3D-EF  ++F33..v66rr) __all__numpyrpatsyr patsy.utilr r r r rrr2r;rErLrOrRrWr]rrpobjectrrrrrrrrrr'rrrs LD & 2&R=I D ;11h@0K6K\AH>&>B#BL9f9x  /6/dB7r