Merge from emacs--rel--22
[bpt/emacs.git] / lisp / calc / calc-nlfit.el
1 ;;; calc-nlfit.el --- nonlinear curve fitting for Calc
2
3 ;; Copyright (C) 2007, 2008 Free Software Foundation, Inc.
4
5 ;; Maintainer: Jay Belanger <jay.p.belanger@gmail.com>
6
7 ;; This file is part of GNU Emacs.
8
9 ;; GNU Emacs is free software; you can redistribute it and/or modify
10 ;; it under the terms of the GNU General Public License as published by
11 ;; the Free Software Foundation; either version 3, or (at your option)
12 ;; any later version.
13
14 ;; GNU Emacs is distributed in the hope that it will be useful,
15 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
16 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 ;; GNU General Public License for more details.
18
19 ;; You should have received a copy of the GNU General Public License
20 ;; along with GNU Emacs; see the file COPYING. If not, write to the
21 ;; Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
22 ;; Boston, MA 02110-1301, USA.
23
24 ;;; Commentary:
25
26 ;; This code uses the Levenberg-Marquardt method, as described in
27 ;; _Numerical Analysis_ by H. R. Schwarz, to fit data to
28 ;; nonlinear curves. Currently, the only the following curves are
29 ;; supported:
30 ;; The logistic S curve, y=a/(1+exp(b*(t-c)))
31 ;; Here, y is usually interpreted as the population of some
32 ;; quantity at time t. So we will think of the data as consisting
33 ;; of quantities q0, q1, ..., qn and their respective times
34 ;; t0, t1, ..., tn.
35
36 ;; The logistic bell curve, y=A*exp(B*(t-C))/(1+exp(B*(t-C)))^2
37 ;; Note that this is the derivative of the formula for the S curve.
38 ;; We get A=-a*b, B=b and C=c. Here, y is interpreted as the rate
39 ;; of growth of a population at time t. So we will think of the
40 ;; data as consisting of rates p0, p1, ..., pn and their
41 ;; respective times t0, t1, ..., tn.
42
43 ;; The Hubbert Linearization, y/x=A*(1-x/B)
44 ;; Here, y is thought of as the rate of growth of a population
45 ;; and x represents the actual population. This is essentially
46 ;; the differential equation describing the actual population.
47
48 ;; The Levenberg-Marquardt method is an iterative process: it takes
49 ;; an initial guess for the parameters and refines them. To get an
50 ;; initial guess for the parameters, we'll use a method described by
51 ;; Luis de Sousa in "Hubbert's Peak Mathematics". The idea is that
52 ;; given quantities Q and the corresponding rates P, they should
53 ;; satisfy P/Q= mQ+a. We can use the parameter a for an
54 ;; approximation for the parameter a in the S curve, and
55 ;; approximations for b and c are found using least squares on the
56 ;; linearization log((a/y)-1) = log(bb) + cc*t of
57 ;; y=a/(1+bb*exp(cc*t)), which is equivalent to the above s curve
58 ;; formula, and then tranlating it to b and c. From this, we can
59 ;; also get approximations for the bell curve parameters.
60
61 ;;; Code:
62
63 (require 'calc-arith)
64 (require 'calcalg3)
65
66 ;; Declare functions which are defined elsewhere.
67 (declare-function calc-get-fit-variables "calcalg3" (nv nc &optional defv defc with-y homog))
68 (declare-function math-map-binop "calcalg3" (binop args1 args2))
69
70 (defun math-nlfit-least-squares (xdata ydata &optional sdata sigmas)
71 "Return the parameters A and B for the best least squares fit y=a+bx."
72 (let* ((n (length xdata))
73 (s2data (if sdata
74 (mapcar 'calcFunc-sqr sdata)
75 (make-list n 1)))
76 (S (if sdata 0 n))
77 (Sx 0)
78 (Sy 0)
79 (Sxx 0)
80 (Sxy 0)
81 D)
82 (while xdata
83 (let ((x (car xdata))
84 (y (car ydata))
85 (s (car s2data)))
86 (setq Sx (math-add Sx (if s (math-div x s) x)))
87 (setq Sy (math-add Sy (if s (math-div y s) y)))
88 (setq Sxx (math-add Sxx (if s (math-div (math-mul x x) s)
89 (math-mul x x))))
90 (setq Sxy (math-add Sxy (if s (math-div (math-mul x y) s)
91 (math-mul x y))))
92 (if sdata
93 (setq S (math-add S (math-div 1 s)))))
94 (setq xdata (cdr xdata))
95 (setq ydata (cdr ydata))
96 (setq s2data (cdr s2data)))
97 (setq D (math-sub (math-mul S Sxx) (math-mul Sx Sx)))
98 (let ((A (math-div (math-sub (math-mul Sxx Sy) (math-mul Sx Sxy)) D))
99 (B (math-div (math-sub (math-mul S Sxy) (math-mul Sx Sy)) D)))
100 (if sigmas
101 (let ((C11 (math-div Sxx D))
102 (C12 (math-neg (math-div Sx D)))
103 (C22 (math-div S D)))
104 (list (list 'sdev A (calcFunc-sqrt C11))
105 (list 'sdev B (calcFunc-sqrt C22))
106 (list 'vec
107 (list 'vec C11 C12)
108 (list 'vec C12 C22))))
109 (list A B)))))
110
111 ;;; The methods described by de Sousa require the cumulative data qdata
112 ;;; and the rates pdata. We will assume that we are given either
113 ;;; qdata and the corresponding times tdata, or pdata and the corresponding
114 ;;; tdata. The following two functions will find pdata or qdata,
115 ;;; given the other..
116
117 ;;; First, given two lists; one of values q0, q1, ..., qn and one of
118 ;;; corresponding times t0, t1, ..., tn; return a list
119 ;;; p0, p1, ..., pn of the rates of change of the qi with respect to t.
120 ;;; p0 is the right hand derivative (q1 - q0)/(t1 - t0).
121 ;;; pn is the left hand derivative (qn - q(n-1))/(tn - t(n-1)).
122 ;;; The other pis are the averages of the two:
123 ;;; (1/2)((qi - q(i-1))/(ti - t(i-1)) + (q(i+1) - qi)/(t(i+1) - ti)).
124
125 (defun math-nlfit-get-rates-from-cumul (tdata qdata)
126 (let ((pdata (list
127 (math-div
128 (math-sub (nth 1 qdata)
129 (nth 0 qdata))
130 (math-sub (nth 1 tdata)
131 (nth 0 tdata))))))
132 (while (> (length qdata) 2)
133 (setq pdata
134 (cons
135 (math-mul
136 '(float 5 -1)
137 (math-add
138 (math-div
139 (math-sub (nth 2 qdata)
140 (nth 1 qdata))
141 (math-sub (nth 2 tdata)
142 (nth 1 tdata)))
143 (math-div
144 (math-sub (nth 1 qdata)
145 (nth 0 qdata))
146 (math-sub (nth 1 tdata)
147 (nth 0 tdata)))))
148 pdata))
149 (setq qdata (cdr qdata)))
150 (setq pdata
151 (cons
152 (math-div
153 (math-sub (nth 1 qdata)
154 (nth 0 qdata))
155 (math-sub (nth 1 tdata)
156 (nth 0 tdata)))
157 pdata))
158 (reverse pdata)))
159
160 ;;; Next, given two lists -- one of rates p0, p1, ..., pn and one of
161 ;;; corresponding times t0, t1, ..., tn -- and an initial values q0,
162 ;;; return a list q0, q1, ..., qn of the cumulative values.
163 ;;; q0 is the initial value given.
164 ;;; For i>0, qi is computed using the trapezoid rule:
165 ;;; qi = q(i-1) + (1/2)(pi + p(i-1))(ti - t(i-1))
166
167 (defun math-nlfit-get-cumul-from-rates (tdata pdata q0)
168 (let* ((qdata (list q0)))
169 (while (cdr pdata)
170 (setq qdata
171 (cons
172 (math-add (car qdata)
173 (math-mul
174 (math-mul
175 '(float 5 -1)
176 (math-add (nth 1 pdata) (nth 0 pdata)))
177 (math-sub (nth 1 tdata)
178 (nth 0 tdata))))
179 qdata))
180 (setq pdata (cdr pdata))
181 (setq tdata (cdr tdata)))
182 (reverse qdata)))
183
184 ;;; Given the qdata, pdata and tdata, find the parameters
185 ;;; a, b and c that fit q = a/(1+b*exp(c*t)).
186 ;;; a is found using the method described by de Sousa.
187 ;;; b and c are found using least squares on the linearization
188 ;;; log((a/q)-1) = log(b) + c*t
189 ;;; In some cases (where the logistic curve may well be the wrong
190 ;;; model), the computed a will be less than or equal to the maximum
191 ;;; value of q in qdata; in which case the above linearization won't work.
192 ;;; In this case, a will be replaced by a number slightly above
193 ;;; the maximum value of q.
194
195 (defun math-nlfit-find-qmax (qdata pdata tdata)
196 (let* ((ratios (math-map-binop 'math-div pdata qdata))
197 (lsdata (math-nlfit-least-squares ratios tdata))
198 (qmax (math-max-list (car qdata) (cdr qdata)))
199 (a (math-neg (math-div (nth 1 lsdata) (nth 0 lsdata)))))
200 (if (math-lessp a qmax)
201 (math-add '(float 5 -1) qmax)
202 a)))
203
204 (defun math-nlfit-find-logistic-parameters (qdata pdata tdata)
205 (let* ((a (math-nlfit-find-qmax qdata pdata tdata))
206 (newqdata
207 (mapcar (lambda (q) (calcFunc-ln (math-sub (math-div a q) 1)))
208 qdata))
209 (bandc (math-nlfit-least-squares tdata newqdata)))
210 (list
211 a
212 (calcFunc-exp (nth 0 bandc))
213 (nth 1 bandc))))
214
215 ;;; Next, given the pdata and tdata, we can find the qdata if we know q0.
216 ;;; We first try to find q0, using the fact that when p takes on its largest
217 ;;; value, q is half of its maximum value. So we'll find the maximum value
218 ;;; of q given various q0, and use bisection to approximate the correct q0.
219
220 ;;; First, given pdata and tdata, find what half of qmax would be if q0=0.
221
222 (defun math-nlfit-find-qmaxhalf (pdata tdata)
223 (let ((pmax (math-max-list (car pdata) (cdr pdata)))
224 (qmh 0))
225 (while (math-lessp (car pdata) pmax)
226 (setq qmh
227 (math-add qmh
228 (math-mul
229 (math-mul
230 '(float 5 -1)
231 (math-add (nth 1 pdata) (nth 0 pdata)))
232 (math-sub (nth 1 tdata)
233 (nth 0 tdata)))))
234 (setq pdata (cdr pdata))
235 (setq tdata (cdr tdata)))
236 qmh))
237
238 ;;; Next, given pdata and tdata, approximate q0.
239
240 (defun math-nlfit-find-q0 (pdata tdata)
241 (let* ((qhalf (math-nlfit-find-qmaxhalf pdata tdata))
242 (q0 (math-mul 2 qhalf))
243 (qdata (math-nlfit-get-cumul-from-rates tdata pdata q0)))
244 (while (math-lessp (math-nlfit-find-qmax
245 (mapcar
246 (lambda (q) (math-add q0 q))
247 qdata)
248 pdata tdata)
249 (math-mul
250 '(float 5 -1)
251 (math-add
252 q0
253 qhalf)))
254 (setq q0 (math-add q0 qhalf)))
255 (let* ((qmin (math-sub q0 qhalf))
256 (qmax q0)
257 (qt (math-nlfit-find-qmax
258 (mapcar
259 (lambda (q) (math-add q0 q))
260 qdata)
261 pdata tdata))
262 (i 0))
263 (while (< i 10)
264 (setq q0 (math-mul '(float 5 -1) (math-add qmin qmax)))
265 (if (math-lessp
266 (math-nlfit-find-qmax
267 (mapcar
268 (lambda (q) (math-add q0 q))
269 qdata)
270 pdata tdata)
271 (math-mul '(float 5 -1) (math-add qhalf q0)))
272 (setq qmin q0)
273 (setq qmax q0))
274 (setq i (1+ i)))
275 (math-mul '(float 5 -1) (math-add qmin qmax)))))
276
277 ;;; To improve the approximations to the parameters, we can use
278 ;;; Marquardt method as described in Schwarz's book.
279
280 ;;; Small numbers used in the Givens algorithm
281 (defvar math-nlfit-delta '(float 1 -8))
282
283 (defvar math-nlfit-epsilon '(float 1 -5))
284
285 ;;; Maximum number of iterations
286 (defvar math-nlfit-max-its 100)
287
288 ;;; Next, we need some functions for dealing with vectors and
289 ;;; matrices. For convenience, we'll work with Emacs lists
290 ;;; as vectors, rather than Calc's vectors.
291
292 (defun math-nlfit-set-elt (vec i x)
293 (setcar (nthcdr (1- i) vec) x))
294
295 (defun math-nlfit-get-elt (vec i)
296 (nth (1- i) vec))
297
298 (defun math-nlfit-make-matrix (i j)
299 (let ((row (make-list j 0))
300 (mat nil)
301 (k 0))
302 (while (< k i)
303 (setq mat (cons (copy-sequence row) mat))
304 (setq k (1+ k)))
305 mat))
306
307 (defun math-nlfit-set-matx-elt (mat i j x)
308 (setcar (nthcdr (1- j) (nth (1- i) mat)) x))
309
310 (defun math-nlfit-get-matx-elt (mat i j)
311 (nth (1- j) (nth (1- i) mat)))
312
313 ;;; For solving the linearized system.
314 ;;; (The Givens method, from Schwarz.)
315
316 (defun math-nlfit-givens (C d)
317 (let* ((C (copy-tree C))
318 (d (copy-tree d))
319 (n (length (car C)))
320 (N (length C))
321 (j 1)
322 (r (make-list N 0))
323 (x (make-list N 0))
324 w
325 gamma
326 sigma
327 rho)
328 (while (<= j n)
329 (let ((i (1+ j)))
330 (while (<= i N)
331 (let ((cij (math-nlfit-get-matx-elt C i j))
332 (cjj (math-nlfit-get-matx-elt C j j)))
333 (when (not (math-equal 0 cij))
334 (if (math-lessp (calcFunc-abs cjj)
335 (math-mul math-nlfit-delta (calcFunc-abs cij)))
336 (setq w (math-neg cij)
337 gamma 0
338 sigma 1
339 rho 1)
340 (setq w (math-mul
341 (calcFunc-sign cjj)
342 (calcFunc-sqrt
343 (math-add
344 (math-mul cjj cjj)
345 (math-mul cij cij))))
346 gamma (math-div cjj w)
347 sigma (math-neg (math-div cij w)))
348 (if (math-lessp (calcFunc-abs sigma) gamma)
349 (setq rho sigma)
350 (setq rho (math-div (calcFunc-sign sigma) gamma))))
351 (setq cjj w
352 cij rho)
353 (math-nlfit-set-matx-elt C j j w)
354 (math-nlfit-set-matx-elt C i j rho)
355 (let ((k (1+ j)))
356 (while (<= k n)
357 (let* ((cjk (math-nlfit-get-matx-elt C j k))
358 (cik (math-nlfit-get-matx-elt C i k))
359 (h (math-sub
360 (math-mul gamma cjk) (math-mul sigma cik))))
361 (setq cik (math-add
362 (math-mul sigma cjk)
363 (math-mul gamma cik)))
364 (setq cjk h)
365 (math-nlfit-set-matx-elt C i k cik)
366 (math-nlfit-set-matx-elt C j k cjk)
367 (setq k (1+ k)))))
368 (let* ((di (math-nlfit-get-elt d i))
369 (dj (math-nlfit-get-elt d j))
370 (h (math-sub
371 (math-mul gamma dj)
372 (math-mul sigma di))))
373 (setq di (math-add
374 (math-mul sigma dj)
375 (math-mul gamma di)))
376 (setq dj h)
377 (math-nlfit-set-elt d i di)
378 (math-nlfit-set-elt d j dj))))
379 (setq i (1+ i))))
380 (setq j (1+ j)))
381 (let ((i n)
382 s)
383 (while (>= i 1)
384 (math-nlfit-set-elt r i 0)
385 (setq s (math-nlfit-get-elt d i))
386 (let ((k (1+ i)))
387 (while (<= k n)
388 (setq s (math-add s (math-mul (math-nlfit-get-matx-elt C i k)
389 (math-nlfit-get-elt x k))))
390 (setq k (1+ k))))
391 (math-nlfit-set-elt x i
392 (math-neg
393 (math-div s
394 (math-nlfit-get-matx-elt C i i))))
395 (setq i (1- i))))
396 (let ((i (1+ n)))
397 (while (<= i N)
398 (math-nlfit-set-elt r i (math-nlfit-get-elt d i))
399 (setq i (1+ i))))
400 (let ((j n))
401 (while (>= j 1)
402 (let ((i N))
403 (while (>= i (1+ j))
404 (setq rho (math-nlfit-get-matx-elt C i j))
405 (if (math-equal rho 1)
406 (setq gamma 0
407 sigma 1)
408 (if (math-lessp (calcFunc-abs rho) 1)
409 (setq sigma rho
410 gamma (calcFunc-sqrt
411 (math-sub 1 (math-mul sigma sigma))))
412 (setq gamma (math-div 1 (calcFunc-abs rho))
413 sigma (math-mul (calcFunc-sign rho)
414 (calcFunc-sqrt
415 (math-sub 1 (math-mul gamma gamma)))))))
416 (let ((ri (math-nlfit-get-elt r i))
417 (rj (math-nlfit-get-elt r j))
418 h)
419 (setq h (math-add (math-mul gamma rj)
420 (math-mul sigma ri)))
421 (setq ri (math-sub
422 (math-mul gamma ri)
423 (math-mul sigma rj)))
424 (setq rj h)
425 (math-nlfit-set-elt r i ri)
426 (math-nlfit-set-elt r j rj))
427 (setq i (1- i))))
428 (setq j (1- j))))
429
430 x))
431
432 (defun math-nlfit-jacobian (grad xlist parms &optional slist)
433 (let ((j nil))
434 (while xlist
435 (let ((row (apply grad (car xlist) parms)))
436 (setq j
437 (cons
438 (if slist
439 (mapcar (lambda (x) (math-div x (car slist))) row)
440 row)
441 j)))
442 (setq slist (cdr slist))
443 (setq xlist (cdr xlist)))
444 (reverse j)))
445
446 (defun math-nlfit-make-ident (l n)
447 (let ((m (math-nlfit-make-matrix n n))
448 (i 1))
449 (while (<= i n)
450 (math-nlfit-set-matx-elt m i i l)
451 (setq i (1+ i)))
452 m))
453
454 (defun math-nlfit-chi-sq (xlist ylist parms fn &optional slist)
455 (let ((cs 0))
456 (while xlist
457 (let ((c
458 (math-sub
459 (apply fn (car xlist) parms)
460 (car ylist))))
461 (if slist
462 (setq c (math-div c (car slist))))
463 (setq cs
464 (math-add cs
465 (math-mul c c))))
466 (setq xlist (cdr xlist))
467 (setq ylist (cdr ylist))
468 (setq slist (cdr slist)))
469 cs))
470
471 (defun math-nlfit-init-lambda (C)
472 (let ((l 0)
473 (n (length (car C)))
474 (N (length C)))
475 (while C
476 (let ((row (car C)))
477 (while row
478 (setq l (math-add l (math-mul (car row) (car row))))
479 (setq row (cdr row))))
480 (setq C (cdr C)))
481 (calcFunc-sqrt (math-div l (math-mul n N)))))
482
483 (defun math-nlfit-make-Ctilda (C l)
484 (let* ((n (length (car C)))
485 (bot (math-nlfit-make-ident l n)))
486 (append C bot)))
487
488 (defun math-nlfit-make-d (fn xdata ydata parms &optional sdata)
489 (let ((d nil))
490 (while xdata
491 (setq d (cons
492 (let ((dd (math-sub (apply fn (car xdata) parms)
493 (car ydata))))
494 (if sdata (math-div dd (car sdata)) dd))
495 d))
496 (setq xdata (cdr xdata))
497 (setq ydata (cdr ydata))
498 (setq sdata (cdr sdata)))
499 (reverse d)))
500
501 (defun math-nlfit-make-dtilda (d n)
502 (append d (make-list n 0)))
503
504 (defun math-nlfit-fit (xlist ylist parms fn grad &optional slist)
505 (let*
506 ((C (math-nlfit-jacobian grad xlist parms slist))
507 (d (math-nlfit-make-d fn xlist ylist parms slist))
508 (chisq (math-nlfit-chi-sq xlist ylist parms fn slist))
509 (lambda (math-nlfit-init-lambda C))
510 (really-done nil)
511 (iters 0))
512 (while (and
513 (not really-done)
514 (< iters math-nlfit-max-its))
515 (setq iters (1+ iters))
516 (let ((done nil))
517 (while (not done)
518 (let* ((Ctilda (math-nlfit-make-Ctilda C lambda))
519 (dtilda (math-nlfit-make-dtilda d (length (car C))))
520 (zeta (math-nlfit-givens Ctilda dtilda))
521 (newparms (math-map-binop 'math-add (copy-tree parms) zeta))
522 (newchisq (math-nlfit-chi-sq xlist ylist newparms fn slist)))
523 (if (math-lessp newchisq chisq)
524 (progn
525 (if (math-lessp
526 (math-div
527 (math-sub chisq newchisq) newchisq) math-nlfit-epsilon)
528 (setq really-done t))
529 (setq lambda (math-div lambda 10))
530 (setq chisq newchisq)
531 (setq parms newparms)
532 (setq done t))
533 (setq lambda (math-mul lambda 10)))))
534 (setq C (math-nlfit-jacobian grad xlist parms slist))
535 (setq d (math-nlfit-make-d fn xlist ylist parms slist))))
536 (list chisq parms)))
537
538 ;;; The functions that describe our models, and their gradients.
539
540 (defun math-nlfit-s-logistic-fn (x a b c)
541 (math-div a (math-add 1 (math-mul b (calcFunc-exp (math-mul c x))))))
542
543 (defun math-nlfit-s-logistic-grad (x a b c)
544 (let* ((ep (calcFunc-exp (math-mul c x)))
545 (d (math-add 1 (math-mul b ep)))
546 (d2 (math-mul d d)))
547 (list
548 (math-div 1 d)
549 (math-neg (math-div (math-mul a ep) d2))
550 (math-neg (math-div (math-mul a (math-mul b (math-mul x ep))) d2)))))
551
552 (defun math-nlfit-b-logistic-fn (x a c d)
553 (let ((ex (calcFunc-exp (math-mul c (math-sub x d)))))
554 (math-div
555 (math-mul a ex)
556 (math-sqr
557 (math-add
558 1 ex)))))
559
560 (defun math-nlfit-b-logistic-grad (x a c d)
561 (let* ((ex (calcFunc-exp (math-mul c (math-sub x d))))
562 (ex1 (math-add 1 ex))
563 (xd (math-sub x d)))
564 (list
565 (math-div
566 ex
567 (math-sqr ex1))
568 (math-sub
569 (math-div
570 (math-mul a (math-mul xd ex))
571 (math-sqr ex1))
572 (math-div
573 (math-mul 2 (math-mul a (math-mul xd (math-sqr ex))))
574 (math-pow ex1 3)))
575 (math-sub
576 (math-div
577 (math-mul 2 (math-mul a (math-mul c (math-sqr ex))))
578 (math-pow ex1 3))
579 (math-div
580 (math-mul a (math-mul c ex))
581 (math-sqr ex1))))))
582
583 ;;; Functions to get the final covariance matrix and the sdevs
584
585 (defun math-nlfit-find-covar (grad xlist pparms)
586 (let ((j nil))
587 (while xlist
588 (setq j (cons (cons 'vec (apply grad (car xlist) pparms)) j))
589 (setq xlist (cdr xlist)))
590 (setq j (cons 'vec (reverse j)))
591 (setq j
592 (math-mul
593 (calcFunc-trn j) j))
594 (calcFunc-inv j)))
595
596 (defun math-nlfit-get-sigmas (grad xlist pparms chisq)
597 (let* ((sgs nil)
598 (covar (math-nlfit-find-covar grad xlist pparms))
599 (n (1- (length covar)))
600 (N (length xlist))
601 (i 1))
602 (when (> N n)
603 (while (<= i n)
604 (setq sgs (cons (calcFunc-sqrt (nth i (nth i covar))) sgs))
605 (setq i (1+ i)))
606 (setq sgs (reverse sgs)))
607 (list sgs covar)))
608
609 ;;; Now the Calc functions
610
611 (defun math-nlfit-s-logistic-params (xdata ydata)
612 (let ((pdata (math-nlfit-get-rates-from-cumul xdata ydata)))
613 (math-nlfit-find-logistic-parameters ydata pdata xdata)))
614
615 (defun math-nlfit-b-logistic-params (xdata ydata)
616 (let* ((q0 (math-nlfit-find-q0 ydata xdata))
617 (qdata (math-nlfit-get-cumul-from-rates xdata ydata q0))
618 (abc (math-nlfit-find-logistic-parameters qdata ydata xdata))
619 (B (nth 1 abc))
620 (C (nth 2 abc))
621 (A (math-neg
622 (math-mul
623 (nth 0 abc)
624 (math-mul B C))))
625 (D (math-neg (math-div (calcFunc-ln B) C)))
626 (A (math-div A B)))
627 (list A C D)))
628
629 ;;; Some functions to turn the parameter lists and variables
630 ;;; into the appropriate functions.
631
632 (defun math-nlfit-s-logistic-solnexpr (pms var)
633 (let ((a (nth 0 pms))
634 (b (nth 1 pms))
635 (c (nth 2 pms)))
636 (list '/ a
637 (list '+
638 1
639 (list '*
640 b
641 (calcFunc-exp
642 (list '*
643 c
644 var)))))))
645
646 (defun math-nlfit-b-logistic-solnexpr (pms var)
647 (let ((a (nth 0 pms))
648 (c (nth 1 pms))
649 (d (nth 2 pms)))
650 (list '/
651 (list '*
652 a
653 (calcFunc-exp
654 (list '*
655 c
656 (list '- var d))))
657 (list '^
658 (list '+
659 1
660 (calcFunc-exp
661 (list '*
662 c
663 (list '- var d))))
664 2))))
665
666 (defun math-nlfit-enter-result (n prefix vals)
667 (setq calc-aborted-prefix prefix)
668 (calc-pop-push-record-list n prefix vals)
669 (calc-handle-whys))
670
671 (defun math-nlfit-fit-curve (fn grad solnexpr initparms &optional sdv)
672 (calc-slow-wrapper
673 (let* ((sdevv (or (eq sdv 'calcFunc-efit) (eq sdv 'calcFunc-xfit)))
674 (calc-display-working-message nil)
675 (data (calc-top 1))
676 (xdata (cdr (car (cdr data))))
677 (ydata (cdr (car (cdr (cdr data)))))
678 (sdata (if (math-contains-sdev-p ydata)
679 (mapcar (lambda (x) (math-get-sdev x t)) ydata)
680 nil))
681 (ydata (mapcar (lambda (x) (math-get-value x)) ydata))
682 (calc-curve-varnames nil)
683 (calc-curve-coefnames nil)
684 (calc-curve-nvars 1)
685 (fitvars (calc-get-fit-variables 1 3))
686 (var (nth 1 calc-curve-varnames))
687 (parms (cdr calc-curve-coefnames))
688 (parmguess
689 (funcall initparms xdata ydata))
690 (fit (math-nlfit-fit xdata ydata parmguess fn grad sdata))
691 (finalparms (nth 1 fit))
692 (sigmacovar
693 (if sdevv
694 (math-nlfit-get-sigmas grad xdata finalparms (nth 0 fit))))
695 (sigmas
696 (if sdevv
697 (nth 0 sigmacovar)))
698 (finalparms
699 (if sigmas
700 (math-map-binop
701 (lambda (x y) (list 'sdev x y)) finalparms sigmas)
702 finalparms))
703 (soln (funcall solnexpr finalparms var)))
704 (let ((calc-fit-to-trail t)
705 (traillist nil))
706 (while parms
707 (setq traillist (cons (list 'calcFunc-eq (car parms) (car finalparms))
708 traillist))
709 (setq finalparms (cdr finalparms))
710 (setq parms (cdr parms)))
711 (setq traillist (calc-normalize (cons 'vec (nreverse traillist))))
712 (cond ((eq sdv 'calcFunc-efit)
713 (math-nlfit-enter-result 1 "efit" soln))
714 ((eq sdv 'calcFunc-xfit)
715 (let (sln)
716 (setq sln
717 (list 'vec
718 soln
719 traillist
720 (nth 1 sigmacovar)
721 '(vec)
722 (nth 0 fit)
723 (let ((n (length xdata))
724 (m (length finalparms)))
725 (if (and sdata (> n m))
726 (calcFunc-utpc (nth 0 fit)
727 (- n m))
728 '(var nan var-nan)))))
729 (math-nlfit-enter-result 1 "xfit" sln)))
730 (t
731 (math-nlfit-enter-result 1 "fit" soln)))
732 (calc-record traillist "parm")))))
733
734 (defun calc-fit-s-shaped-logistic-curve (arg)
735 (interactive "P")
736 (math-nlfit-fit-curve 'math-nlfit-s-logistic-fn
737 'math-nlfit-s-logistic-grad
738 'math-nlfit-s-logistic-solnexpr
739 'math-nlfit-s-logistic-params
740 arg))
741
742 (defun calc-fit-bell-shaped-logistic-curve (arg)
743 (interactive "P")
744 (math-nlfit-fit-curve 'math-nlfit-b-logistic-fn
745 'math-nlfit-b-logistic-grad
746 'math-nlfit-b-logistic-solnexpr
747 'math-nlfit-b-logistic-params
748 arg))
749
750 (defun calc-fit-hubbert-linear-curve (&optional sdv)
751 (calc-slow-wrapper
752 (let* ((sdevv (or (eq sdv 'calcFunc-efit) (eq sdv 'calcFunc-xfit)))
753 (calc-display-working-message nil)
754 (data (calc-top 1))
755 (qdata (cdr (car (cdr data))))
756 (pdata (cdr (car (cdr (cdr data)))))
757 (sdata (if (math-contains-sdev-p pdata)
758 (mapcar (lambda (x) (math-get-sdev x t)) pdata)
759 nil))
760 (pdata (mapcar (lambda (x) (math-get-value x)) pdata))
761 (poverqdata (math-map-binop 'math-div pdata qdata))
762 (parmvals (math-nlfit-least-squares qdata poverqdata sdata sdevv))
763 (finalparms (list (nth 0 parmvals)
764 (math-neg
765 (math-div (nth 0 parmvals)
766 (nth 1 parmvals)))))
767 (calc-curve-varnames nil)
768 (calc-curve-coefnames nil)
769 (calc-curve-nvars 1)
770 (fitvars (calc-get-fit-variables 1 2))
771 (var (nth 1 calc-curve-varnames))
772 (parms (cdr calc-curve-coefnames))
773 (soln (list '* (nth 0 finalparms)
774 (list '- 1
775 (list '/ var (nth 1 finalparms))))))
776 (let ((calc-fit-to-trail t)
777 (traillist nil))
778 (setq traillist
779 (list 'vec
780 (list 'calcFunc-eq (nth 0 parms) (nth 0 finalparms))
781 (list 'calcFunc-eq (nth 1 parms) (nth 1 finalparms))))
782 (cond ((eq sdv 'calcFunc-efit)
783 (math-nlfit-enter-result 1 "efit" soln))
784 ((eq sdv 'calcFunc-xfit)
785 (let (sln
786 (chisq
787 (math-nlfit-chi-sq
788 qdata poverqdata
789 (list (nth 1 (nth 0 finalparms))
790 (nth 1 (nth 1 finalparms)))
791 (lambda (x a b)
792 (math-mul a
793 (math-sub
794 1
795 (math-div x b))))
796 sdata)))
797 (setq sln
798 (list 'vec
799 soln
800 traillist
801 (nth 2 parmvals)
802 (list
803 'vec
804 '(calcFunc-fitdummy 1)
805 (list 'calcFunc-neg
806 (list '/
807 '(calcFunc-fitdummy 1)
808 '(calcFunc-fitdummy 2))))
809 chisq
810 (let ((n (length qdata)))
811 (if (and sdata (> n 2))
812 (calcFunc-utpc
813 chisq
814 (- n 2))
815 '(var nan var-nan)))))
816 (math-nlfit-enter-result 1 "xfit" sln)))
817 (t
818 (math-nlfit-enter-result 1 "fit" soln)))
819 (calc-record traillist "parm")))))
820
821 (provide 'calc-nlfit)
822
823 ;; arch-tag: 6eba3cd6-f48b-4a84-8174-10c15a024928