>h'DSrSSKrSSKJrJrJr SSKJr "SS5r g)aJ Created on Sat Oct 01 20:20:16 2011 Author: Josef Perktold License: BSD-3 TODO: check orientation, size and alpha should be increasing for interp1d, but what is alpha? can be either sf or cdf probability change it to use one consistent notation check: instead of bound checking I could use the fill-value of the interpolators N)interp1dinterp2dRbf)cache_readonlycn\rSrSrSrS Sjr\S5r\S5r\S5r Sr S r S r S r S rg) TableDista Distribution, critical values and p-values from tables currently only 1 extra parameter, e.g. sample size Parameters ---------- alpha : array_like, 1d probabiliy in the table, could be either sf (right tail) or cdf (left tail) size : array_like, 1d The sample sizes for the table crit_table : array_like, 2d The sample sizes in the table array with critical values for sample size in rows and probability in columns asymptotic : callable, optional Callable function with the form fn(nobs) that returns len(alpha) critical values where the critical value in position i corresponds to alpha[i] min_nobs : int, optional Minimum number of observations to use the asymptotic distribution. If not provided, uses max(size). max_nobs : int, optional Maximum number of observations to use the tabular distribution. If not provided, uses max(size) Notes ----- size and alpha must be sorted and increasing. If both min_nobs and max_nobs are provided, then the critical values from the tabular distribution and the asymptotic distribution are linearly blended using the formula :math:`w cv_a + (1-w) cv_t` where the weight is :math:`w = (n - a_{min}) / (a_{max} - a_{min})`. This ensures the transition between the tabular and the asymptotic critical values is continuous. If these are not provided, then the asymptotic critical value is used for nobs > max(size). Nc[R"U5UlURRS:wa [ S5e[R "UR5S:*R 5(a [ S5e[R"U5UlURRS:wa [ S5e[R "UR5S:*R 5(a [ S5eURRS:Xa7[R "U5S:*R 5(a [ S5e[R"U5UlURRURRSURRS4:wa [ S5e[U5Ul [R"[R "URS5R55UlURS:aURSS2SS/4UlOURSS2SS/4UlSUl[#U5=oplUb?U"UR$S-5n[U5[U5:wa [ S 5eX@lUcUOUUlUcUOUUlUR.U:a [ S 5eUR0U:a [ S 5eg![&a0n [)U 5"SR+U R,SS 95eSn A ff=f) Nzalpha is not 1drzalpha is not sortedzsize is not 1dzsize is not sortedz1crit_table must have shape(len(size), len(alpha))zICalling asymptotic(self.size+1) failed. The error message was: {err_msg})err_msgz,asymptotic does not return len(alpha) valueszmin_nobs > max(size)zmax_nobs > max(size))npasarrayalphandim ValueErrordiffanysize crit_tableshapelenn_alphasignmeansigncrit critv_bounds asymptoticmaxmax_size Exceptiontypeformatargsmin_nobsmax_nobs) selfrrrrr$r%rcvexcs nC:\Users\julio\OneDrive\Documentos\Trabajo\Ideas Frescas\venv\Lib\site-packages\statsmodels/stats/tabledist.py__init__TableDist.__init__?sZZ& ::??a ./ /ggdjj!Q& + + - -23 3JJt$ 99>>Q -. .ggdii A% * * , ,12 2 99>>Q !#((** !677**Z0 ?? TYY__Q%79I9I!9L$M M78 85z  ; @ @ BC ==1  $Aq6 :D  $Aq6 :D #&t9,=  ! M  12 2w#e*$ "*++(O$,$4( $,$4( ==8 #34 4 ==8 #34 4 $ M3i!006sxx{0KMM Ms=L M+L??Mc [UR5Vs/sH*n[URURSS2U45PM, nnU$s snfN)rangerrrr)r&ipolyns r)r0TableDist.polynosO  -/-Q$))T__QT%:;- / /s1A c\[URURUR5nU$r-)rrrr)r&poly2ds r)r3TableDist.poly2dus#$))TZZA c[R"URR[5UR 5up[ UR5UR5URRR5SS9nU$)Nlinear)function) r meshgridrastypefloatrrravelrT)r&xsxapolyrbfs r)r@TableDist.polyrbf|s^TYY--e4djjAbhhj"((*doo.?.?.E.E.G')r5cXR:a+URbURU5nU$[S5eURVs/sH o3"U5PM nnXR:aRXR- UR UR- - n[ SU5nURU5nXE-SU- U--nU$s snf)a Rows of the table, linearly interpolated for given sample size Parameters ---------- n : float sample size, second parameter of the table Returns ------- critv : ndarray, 1d critical values (ppf) corresponding to a row of the table Notes ----- This is used in two step interpolation, or if we want to know the critical values for all alphas for any sample size that we can obtain through interpolation z?n is above max(size) and no asymptotic distribtuion is providedg?r )rrrr0r$r%min)r&nr'pwa_cvs r) _critvalsTableDist._critvalss( }} *__Q' !"<=="&,A1Q4B,== &4==4==+HIQKq)XQ" , -s B?cURU5nURnURS:aUSSS2USSS2pC[R"U5S:Xa.XS:aUS$XS:aUS$[ X45"U5S$XS:nXS:n[R "XV5)n[R[R"UR5-nUSX'USX'[ X45"X5X'U$)a Find pvalues by interpolation, either cdf(x) Returns extreme probabilities, 0.001 and 0.2, for out of range Parameters ---------- x : array_like observed value, assumed to follow the distribution in the table n : float sample size, second parameter of the table Returns ------- prob : array_like This is the probability for each value of x, the p-value in underlying distribution is for a statistical test. r Nr) rHrrr rr logical_ornanonesr) r&xrDcritvrcond_low cond_high cond_interiorprobss r)probTableDist.probs&q!  ==1  2;dd 5 771:?8|Qx2YRy E)!,R0 0!H H2YI]]8??MFFRWWQWW--E#AhEO#BiEO#+E#9!:J#KE Lr5c[R"U5nURnURU5nXS:nXS:n[R"XV5nUR S:Xa(U(a[ X45"U5$[R$[R[R"UR5-n[ X45"X5X'U$)ao Returns interpolated quantiles, similar to ppf or isf use two sequential 1d interpolation, first by n then by prob Parameters ---------- prob : array_like probabilities corresponding to the definition of table columns n : int or float sample size, second parameter of the table Returns ------- ppf : array_like critical values with same shape as prob rrKr ) r rrrHrMrrrNrOr) r&rVrDrrQ cond_ilow cond_ihighrTquantiles r)critTableDist.crits$zz$ q!!H_ 2Y&  i<  99>-d33vv 66BGGDJJ//"*5"89L"Mr5c[R"U5nURnXS:nXS:n[R"XE5nURS:Xa(U(aUR X!5$[R $[R [R"UR5-nUR X!U5Xv'U$)a Returns interpolated quantiles, similar to ppf or isf uses Rbf to interpolate critical values as function of `prob` and `n` Parameters ---------- prob : array_like probabilities corresponding to the definition of table columns n : int or float sample size, second parameter of the table Returns ------- ppf : array_like critical values with same shape as prob, returns nan for arguments that are outside of the table bounds rrKr ) r rrrMrr@rNrOr)r&rVrDrrYrZrTr[s r)crit3TableDist.crit3s&zz$ !H_ 2Y&  i<  99>||A,,vv 66BGGDJJ//"&,,q}2E"Fr5) rrrrr%rr$rrr)NNN)__name__ __module__ __qualname____firstlineno____doc__r*rr0r3r@rHrVr\r___static_attributes__rLr5r)rrsj'R<@)-.5`   "H,\%N&r5r) renumpyr scipy.interpolaterrrstatsmodels.tools.decoratorsrrrLr5r)rjs$ 557MMr5