1 | """ |
---|
2 | This software was developed by the University of Tennessee as part of the |
---|
3 | Distributed Data Analysis of Neutron Scattering Experiments (DANSE) |
---|
4 | project funded by the US National Science Foundation. |
---|
5 | |
---|
6 | See the license text in license.txt |
---|
7 | |
---|
8 | copyright 2010, University of Tennessee |
---|
9 | """ |
---|
10 | |
---|
11 | import os.path |
---|
12 | import unittest |
---|
13 | import math |
---|
14 | import numpy as np |
---|
15 | from sas.sascalc.dataloader.loader import Loader |
---|
16 | from sas.sascalc.dataloader.data_info import Data1D |
---|
17 | |
---|
18 | from sas.sascalc.invariant import invariant |
---|
19 | |
---|
20 | |
---|
21 | def find(filename): |
---|
22 | return os.path.join(os.path.dirname(__file__), filename) |
---|
23 | |
---|
24 | |
---|
25 | class TestLinearFit(unittest.TestCase): |
---|
26 | """ |
---|
27 | Test Line fit |
---|
28 | """ |
---|
29 | def setUp(self): |
---|
30 | x = np.asarray([1.,2.,3.,4.,5.,6.,7.,8.,9.]) |
---|
31 | y = np.asarray([1.,2.,3.,4.,5.,6.,7.,8.,9.]) |
---|
32 | dy = y/10.0 |
---|
33 | |
---|
34 | self.data = Data1D(x=x,y=y,dy=dy) |
---|
35 | |
---|
36 | def test_fit_linear_data(self): |
---|
37 | """ |
---|
38 | Simple linear fit |
---|
39 | """ |
---|
40 | |
---|
41 | # Create invariant object. Background and scale left as defaults. |
---|
42 | fit = invariant.Extrapolator(data=self.data) |
---|
43 | #a,b = fit.fit() |
---|
44 | p, dp = fit.fit() |
---|
45 | |
---|
46 | # Test results |
---|
47 | self.assertAlmostEquals(p[0], 1.0, 5) |
---|
48 | self.assertAlmostEquals(p[1], 0.0, 5) |
---|
49 | |
---|
50 | def test_fit_linear_data_with_noise(self): |
---|
51 | """ |
---|
52 | Simple linear fit with noise |
---|
53 | """ |
---|
54 | import random, math |
---|
55 | |
---|
56 | for i in range(len(self.data.y)): |
---|
57 | self.data.y[i] = self.data.y[i]+.1*(random.random()-0.5) |
---|
58 | |
---|
59 | # Create invariant object. Background and scale left as defaults. |
---|
60 | fit = invariant.Extrapolator(data=self.data) |
---|
61 | p, dp = fit.fit() |
---|
62 | |
---|
63 | # Test results |
---|
64 | self.assertTrue(math.fabs(p[0]-1.0)<0.05) |
---|
65 | self.assertTrue(math.fabs(p[1])<0.1) |
---|
66 | |
---|
67 | def test_fit_with_fixed_parameter(self): |
---|
68 | """ |
---|
69 | Linear fit for y=ax+b where a is fixed. |
---|
70 | """ |
---|
71 | # Create invariant object. Background and scale left as defaults. |
---|
72 | fit = invariant.Extrapolator(data=self.data) |
---|
73 | p, dp = fit.fit(power=-1.0) |
---|
74 | |
---|
75 | # Test results |
---|
76 | self.assertAlmostEquals(p[0], 1.0, 5) |
---|
77 | self.assertAlmostEquals(p[1], 0.0, 5) |
---|
78 | |
---|
79 | def test_fit_linear_data_with_noise_and_fixed_par(self): |
---|
80 | """ |
---|
81 | Simple linear fit with noise |
---|
82 | """ |
---|
83 | import random, math |
---|
84 | |
---|
85 | for i in range(len(self.data.y)): |
---|
86 | self.data.y[i] = self.data.y[i]+.1*(random.random()-0.5) |
---|
87 | |
---|
88 | # Create invariant object. Background and scale left as defaults. |
---|
89 | fit = invariant.Extrapolator(data=self.data) |
---|
90 | p, dp = fit.fit(power=-1.0) |
---|
91 | |
---|
92 | # Test results |
---|
93 | self.assertTrue(math.fabs(p[0]-1.0)<0.05) |
---|
94 | self.assertTrue(math.fabs(p[1])<0.1) |
---|
95 | |
---|
96 | |
---|
97 | |
---|
98 | class TestInvariantCalculator(unittest.TestCase): |
---|
99 | """ |
---|
100 | Test main functionality of the Invariant calculator |
---|
101 | """ |
---|
102 | def setUp(self): |
---|
103 | data = Loader().load(find("latex_smeared_slit.xml")) |
---|
104 | self.data = data[0] |
---|
105 | self.data.dxl = None |
---|
106 | |
---|
107 | def test_initial_data_processing(self): |
---|
108 | """ |
---|
109 | Test whether the background and scale are handled properly |
---|
110 | when creating an InvariantCalculator object |
---|
111 | """ |
---|
112 | length = len(self.data.x) |
---|
113 | self.assertEqual(length, len(self.data.y)) |
---|
114 | inv = invariant.InvariantCalculator(self.data) |
---|
115 | |
---|
116 | self.assertEqual(length, len(inv._data.x)) |
---|
117 | self.assertEqual(inv._data.x[0], self.data.x[0]) |
---|
118 | |
---|
119 | # Now the same thing with a background value |
---|
120 | bck = 0.1 |
---|
121 | inv = invariant.InvariantCalculator(self.data, background=bck) |
---|
122 | self.assertEqual(inv._background, bck) |
---|
123 | |
---|
124 | self.assertEqual(length, len(inv._data.x)) |
---|
125 | self.assertEqual(inv._data.y[0]+bck, self.data.y[0]) |
---|
126 | |
---|
127 | # Now the same thing with a scale value |
---|
128 | scale = 0.1 |
---|
129 | inv = invariant.InvariantCalculator(self.data, scale=scale) |
---|
130 | self.assertEqual(inv._scale, scale) |
---|
131 | |
---|
132 | self.assertEqual(length, len(inv._data.x)) |
---|
133 | self.assertAlmostEqual(inv._data.y[0]/scale, self.data.y[0],7) |
---|
134 | |
---|
135 | |
---|
136 | def test_incompatible_data_class(self): |
---|
137 | """ |
---|
138 | Check that only classes that inherit from Data1D are allowed as data. |
---|
139 | """ |
---|
140 | class Incompatible(): |
---|
141 | pass |
---|
142 | self.assertRaises(ValueError, invariant.InvariantCalculator, Incompatible()) |
---|
143 | |
---|
144 | def test_error_treatment(self): |
---|
145 | x = np.asarray(np.asarray([0,1,2,3])) |
---|
146 | y = np.asarray(np.asarray([1,1,1,1])) |
---|
147 | |
---|
148 | # These are all the values of the dy array that would cause |
---|
149 | # us to set all dy values to 1.0 at __init__ time. |
---|
150 | dy_list = [ [], None, [0,0,0,0] ] |
---|
151 | |
---|
152 | for dy in dy_list: |
---|
153 | data = Data1D(x=x, y=y, dy=dy) |
---|
154 | inv = invariant.InvariantCalculator(data) |
---|
155 | self.assertEqual(len(inv._data.x), len(inv._data.dy)) |
---|
156 | self.assertEqual(len(inv._data.dy), 4) |
---|
157 | for i in range(4): |
---|
158 | self.assertEqual(inv._data.dy[i],1) |
---|
159 | |
---|
160 | def test_qstar_low_q_guinier(self): |
---|
161 | """ |
---|
162 | Test low-q extrapolation with a Guinier |
---|
163 | """ |
---|
164 | inv = invariant.InvariantCalculator(self.data) |
---|
165 | |
---|
166 | # Basic sanity check |
---|
167 | _qstar = inv.get_qstar() |
---|
168 | qstar, dqstar = inv.get_qstar_with_error() |
---|
169 | self.assertEqual(qstar, _qstar) |
---|
170 | |
---|
171 | # Low-Q Extrapolation |
---|
172 | # Check that the returned invariant is what we expect given |
---|
173 | # the result we got without extrapolation |
---|
174 | inv.set_extrapolation('low', npts=10, function='guinier') |
---|
175 | qs_extr, dqs_extr = inv.get_qstar_with_error('low') |
---|
176 | delta_qs_extr, delta_dqs_extr = inv.get_qstar_low() |
---|
177 | |
---|
178 | self.assertEqual(qs_extr, _qstar+delta_qs_extr) |
---|
179 | self.assertEqual(dqs_extr, math.sqrt(dqstar*dqstar + delta_dqs_extr*delta_dqs_extr)) |
---|
180 | |
---|
181 | # We don't expect the extrapolated invariant to be very far from the |
---|
182 | # result without extrapolation. Let's test for a result within 10%. |
---|
183 | self.assertTrue(math.fabs(qs_extr-qstar)/qstar<0.1) |
---|
184 | |
---|
185 | # Check that the two results are consistent within errors |
---|
186 | # Note that the error on the extrapolated value takes into account |
---|
187 | # a systematic error for the fact that we may not know the shape of I(q) at low Q. |
---|
188 | self.assertTrue(math.fabs(qs_extr-qstar)<dqs_extr) |
---|
189 | |
---|
190 | def test_qstar_low_q_power_law(self): |
---|
191 | """ |
---|
192 | Test low-q extrapolation with a power law |
---|
193 | """ |
---|
194 | inv = invariant.InvariantCalculator(self.data) |
---|
195 | |
---|
196 | # Basic sanity check |
---|
197 | _qstar = inv.get_qstar() |
---|
198 | qstar, dqstar = inv.get_qstar_with_error() |
---|
199 | self.assertEqual(qstar, _qstar) |
---|
200 | |
---|
201 | # Low-Q Extrapolation |
---|
202 | # Check that the returned invariant is what we expect given |
---|
203 | inv.set_extrapolation('low', npts=10, function='power_law') |
---|
204 | qs_extr, dqs_extr = inv.get_qstar_with_error('low') |
---|
205 | delta_qs_extr, delta_dqs_extr = inv.get_qstar_low() |
---|
206 | |
---|
207 | # A fit using SasView gives 0.0655 for the value of the exponent |
---|
208 | self.assertAlmostEqual(inv._low_extrapolation_function.power, 0.0655, 3) |
---|
209 | |
---|
210 | if False: |
---|
211 | npts = len(inv._data.x)-1 |
---|
212 | import matplotlib.pyplot as plt |
---|
213 | plt.loglog(inv._data.x[:npts], inv._data.y[:npts], 'o', label='Original data', markersize=10) |
---|
214 | plt.loglog(inv._data.x[:npts], inv._low_extrapolation_function.evaluate_model(inv._data.x[:npts]), 'r', label='Fitted line') |
---|
215 | plt.legend() |
---|
216 | plt.show() |
---|
217 | |
---|
218 | self.assertEqual(qs_extr, _qstar+delta_qs_extr) |
---|
219 | self.assertAlmostEqual(dqs_extr, math.sqrt(dqstar*dqstar + delta_dqs_extr*delta_dqs_extr), 15) |
---|
220 | |
---|
221 | # We don't expect the extrapolated invariant to be very far from the |
---|
222 | # result without extrapolation. Let's test for a result within 10%. |
---|
223 | self.assertTrue(math.fabs(qs_extr-qstar)/qstar<0.1) |
---|
224 | |
---|
225 | # Check that the two results are consistent within errors |
---|
226 | # Note that the error on the extrapolated value takes into account |
---|
227 | # a systematic error for the fact that we may not know the shape of I(q) at low Q. |
---|
228 | self.assertTrue(math.fabs(qs_extr-qstar)<dqs_extr) |
---|
229 | |
---|
230 | def test_qstar_high_q(self): |
---|
231 | """ |
---|
232 | Test high-q extrapolation |
---|
233 | """ |
---|
234 | inv = invariant.InvariantCalculator(self.data) |
---|
235 | |
---|
236 | # Basic sanity check |
---|
237 | _qstar = inv.get_qstar() |
---|
238 | qstar, dqstar = inv.get_qstar_with_error() |
---|
239 | self.assertEqual(qstar, _qstar) |
---|
240 | |
---|
241 | # High-Q Extrapolation |
---|
242 | # Check that the returned invariant is what we expect given |
---|
243 | # the result we got without extrapolation |
---|
244 | inv.set_extrapolation('high', npts=20, function='power_law') |
---|
245 | qs_extr, dqs_extr = inv.get_qstar_with_error('high') |
---|
246 | delta_qs_extr, delta_dqs_extr = inv.get_qstar_high() |
---|
247 | |
---|
248 | # From previous analysis using SasView, we expect an exponent of about 3 |
---|
249 | self.assertTrue(math.fabs(inv._high_extrapolation_function.power-3)<0.1) |
---|
250 | |
---|
251 | self.assertEqual(qs_extr, _qstar+delta_qs_extr) |
---|
252 | self.assertAlmostEqual(dqs_extr, math.sqrt(dqstar*dqstar + delta_dqs_extr*delta_dqs_extr), 10) |
---|
253 | |
---|
254 | # We don't expect the extrapolated invariant to be very far from the |
---|
255 | # result without extrapolation. Let's test for a result within 10%. |
---|
256 | #TODO: verify whether this test really makes sense |
---|
257 | #self.assertTrue(math.fabs(qs_extr-qstar)/qstar<0.1) |
---|
258 | |
---|
259 | # Check that the two results are consistent within errors |
---|
260 | self.assertTrue(math.fabs(qs_extr-qstar)<dqs_extr) |
---|
261 | |
---|
262 | def test_qstar_full_q(self): |
---|
263 | """ |
---|
264 | Test high-q extrapolation |
---|
265 | """ |
---|
266 | inv = invariant.InvariantCalculator(self.data) |
---|
267 | |
---|
268 | # Basic sanity check |
---|
269 | _qstar = inv.get_qstar() |
---|
270 | qstar, dqstar = inv.get_qstar_with_error() |
---|
271 | self.assertEqual(qstar, _qstar) |
---|
272 | |
---|
273 | # High-Q Extrapolation |
---|
274 | # Check that the returned invariant is what we expect given |
---|
275 | # the result we got without extrapolation |
---|
276 | inv.set_extrapolation('low', npts=10, function='guinier') |
---|
277 | inv.set_extrapolation('high', npts=20, function='power_law') |
---|
278 | qs_extr, dqs_extr = inv.get_qstar_with_error('both') |
---|
279 | delta_qs_low, delta_dqs_low = inv.get_qstar_low() |
---|
280 | delta_qs_hi, delta_dqs_hi = inv.get_qstar_high() |
---|
281 | |
---|
282 | self.assertAlmostEqual(qs_extr, _qstar+delta_qs_low+delta_qs_hi, 8) |
---|
283 | self.assertAlmostEqual(dqs_extr, math.sqrt(dqstar*dqstar + delta_dqs_low*delta_dqs_low \ |
---|
284 | + delta_dqs_hi*delta_dqs_hi), 8) |
---|
285 | |
---|
286 | # We don't expect the extrapolated invariant to be very far from the |
---|
287 | # result without extrapolation. Let's test for a result within 10%. |
---|
288 | #TODO: verify whether this test really makes sense |
---|
289 | #self.assertTrue(math.fabs(qs_extr-qstar)/qstar<0.1) |
---|
290 | |
---|
291 | # Check that the two results are consistent within errors |
---|
292 | self.assertTrue(math.fabs(qs_extr-qstar)<dqs_extr) |
---|
293 | |
---|
294 | def _check_values(to_check, reference, tolerance=0.05): |
---|
295 | self.assertTrue( math.fabs(to_check-reference)/reference < tolerance, msg="Tested value = "+str(to_check) ) |
---|
296 | |
---|
297 | # The following values should be replaced by values pulled from IGOR |
---|
298 | # Volume Fraction: |
---|
299 | v, dv = inv.get_volume_fraction_with_error(1, None) |
---|
300 | _check_values(v, 1.88737914186e-15) |
---|
301 | |
---|
302 | v_l, dv_l = inv.get_volume_fraction_with_error(1, 'low') |
---|
303 | _check_values(v_l, 1.94289029309e-15) |
---|
304 | |
---|
305 | v_h, dv_h = inv.get_volume_fraction_with_error(1, 'high') |
---|
306 | _check_values(v_h, 6.99440505514e-15) |
---|
307 | |
---|
308 | v_b, dv_b = inv.get_volume_fraction_with_error(1, 'both') |
---|
309 | _check_values(v_b, 6.99440505514e-15) |
---|
310 | |
---|
311 | # Specific Surface: |
---|
312 | s, ds = inv.get_surface_with_error(1, 1, None) |
---|
313 | _check_values(s, 3.1603095786e-09) |
---|
314 | |
---|
315 | s_l, ds_l = inv.get_surface_with_error(1, 1, 'low') |
---|
316 | _check_values(s_l, 3.1603095786e-09) |
---|
317 | |
---|
318 | s_h, ds_h = inv.get_surface_with_error(1, 1, 'high') |
---|
319 | _check_values(s_h, 3.1603095786e-09) |
---|
320 | |
---|
321 | s_b, ds_b = inv.get_surface_with_error(1, 1, 'both') |
---|
322 | _check_values(s_b, 3.1603095786e-09) |
---|
323 | |
---|
324 | |
---|
325 | def test_bad_parameter_name(self): |
---|
326 | """ |
---|
327 | The set_extrapolation method checks that the name of the extrapolation |
---|
328 | function and the name of the q-range to extrapolate (high/low) is |
---|
329 | recognized. |
---|
330 | """ |
---|
331 | inv = invariant.InvariantCalculator(self.data) |
---|
332 | self.assertRaises(ValueError, inv.set_extrapolation, 'low', npts=4, function='not_a_name') |
---|
333 | self.assertRaises(ValueError, inv.set_extrapolation, 'not_a_range', npts=4, function='guinier') |
---|
334 | self.assertRaises(ValueError, inv.set_extrapolation, 'high', npts=4, function='guinier') |
---|
335 | |
---|
336 | |
---|
337 | class TestGuinierExtrapolation(unittest.TestCase): |
---|
338 | """ |
---|
339 | Generate a Guinier distribution and verify that the extrapolation |
---|
340 | produce the correct ditribution. |
---|
341 | """ |
---|
342 | |
---|
343 | def setUp(self): |
---|
344 | """ |
---|
345 | Generate a Guinier distribution. After extrapolating, we will |
---|
346 | verify that we obtain the scale and rg parameters |
---|
347 | """ |
---|
348 | self.scale = 1.5 |
---|
349 | self.rg = 30.0 |
---|
350 | x = np.arange(0.0001, 0.1, 0.0001) |
---|
351 | y = np.asarray([self.scale * math.exp( -(q*self.rg)**2 / 3.0 ) for q in x]) |
---|
352 | dy = y*.1 |
---|
353 | self.data = Data1D(x=x, y=y, dy=dy) |
---|
354 | |
---|
355 | def test_low_q(self): |
---|
356 | """ |
---|
357 | Invariant with low-Q extrapolation |
---|
358 | """ |
---|
359 | # Create invariant object. Background and scale left as defaults. |
---|
360 | inv = invariant.InvariantCalculator(data=self.data) |
---|
361 | # Set the extrapolation parameters for the low-Q range |
---|
362 | inv.set_extrapolation(range='low', npts=20, function='guinier') |
---|
363 | |
---|
364 | self.assertEqual(inv._low_extrapolation_npts, 20) |
---|
365 | self.assertEqual(inv._low_extrapolation_function.__class__, invariant.Guinier) |
---|
366 | |
---|
367 | # Data boundaries for fiiting |
---|
368 | qmin = inv._data.x[0] |
---|
369 | qmax = inv._data.x[inv._low_extrapolation_npts - 1] |
---|
370 | |
---|
371 | # Extrapolate the low-Q data |
---|
372 | inv._fit(model=inv._low_extrapolation_function, |
---|
373 | qmin=qmin, |
---|
374 | qmax=qmax, |
---|
375 | power=inv._low_extrapolation_power) |
---|
376 | self.assertAlmostEqual(self.scale, inv._low_extrapolation_function.scale, 6) |
---|
377 | self.assertAlmostEqual(self.rg, inv._low_extrapolation_function.radius, 6) |
---|
378 | |
---|
379 | |
---|
380 | class TestPowerLawExtrapolation(unittest.TestCase): |
---|
381 | """ |
---|
382 | Generate a power law distribution and verify that the extrapolation |
---|
383 | produce the correct ditribution. |
---|
384 | """ |
---|
385 | |
---|
386 | def setUp(self): |
---|
387 | """ |
---|
388 | Generate a power law distribution. After extrapolating, we will |
---|
389 | verify that we obtain the scale and m parameters |
---|
390 | """ |
---|
391 | self.scale = 1.5 |
---|
392 | self.m = 3.0 |
---|
393 | x = np.arange(0.0001, 0.1, 0.0001) |
---|
394 | y = np.asarray([self.scale * math.pow(q ,-1.0*self.m) for q in x]) |
---|
395 | dy = y*.1 |
---|
396 | self.data = Data1D(x=x, y=y, dy=dy) |
---|
397 | |
---|
398 | def test_low_q(self): |
---|
399 | """ |
---|
400 | Invariant with low-Q extrapolation |
---|
401 | """ |
---|
402 | # Create invariant object. Background and scale left as defaults. |
---|
403 | inv = invariant.InvariantCalculator(data=self.data) |
---|
404 | # Set the extrapolation parameters for the low-Q range |
---|
405 | inv.set_extrapolation(range='low', npts=20, function='power_law') |
---|
406 | |
---|
407 | self.assertEqual(inv._low_extrapolation_npts, 20) |
---|
408 | self.assertEqual(inv._low_extrapolation_function.__class__, invariant.PowerLaw) |
---|
409 | |
---|
410 | # Data boundaries for fitting |
---|
411 | qmin = inv._data.x[0] |
---|
412 | qmax = inv._data.x[inv._low_extrapolation_npts - 1] |
---|
413 | |
---|
414 | # Extrapolate the low-Q data |
---|
415 | inv._fit(model=inv._low_extrapolation_function, |
---|
416 | qmin=qmin, |
---|
417 | qmax=qmax, |
---|
418 | power=inv._low_extrapolation_power) |
---|
419 | |
---|
420 | self.assertAlmostEqual(self.scale, inv._low_extrapolation_function.scale, 6) |
---|
421 | self.assertAlmostEqual(self.m, inv._low_extrapolation_function.power, 6) |
---|
422 | |
---|
423 | class TestLinearization(unittest.TestCase): |
---|
424 | |
---|
425 | def test_guinier_incompatible_length(self): |
---|
426 | g = invariant.Guinier() |
---|
427 | data_in = Data1D(x=[1], y=[1,2], dy=None) |
---|
428 | self.assertRaises(AssertionError, g.linearize_data, data_in) |
---|
429 | data_in = Data1D(x=[1,1], y=[1,2], dy=[1]) |
---|
430 | self.assertRaises(AssertionError, g.linearize_data, data_in) |
---|
431 | |
---|
432 | def test_linearization(self): |
---|
433 | """ |
---|
434 | Check that the linearization process filters out points |
---|
435 | that can't be transformed |
---|
436 | """ |
---|
437 | x = np.asarray(np.asarray([0,1,2,3])) |
---|
438 | y = np.asarray(np.asarray([1,1,1,1])) |
---|
439 | g = invariant.Guinier() |
---|
440 | data_in = Data1D(x=x, y=y) |
---|
441 | data_out = g.linearize_data(data_in) |
---|
442 | x_out, y_out, dy_out = data_out.x, data_out.y, data_out.dy |
---|
443 | self.assertEqual(len(x_out), 3) |
---|
444 | self.assertEqual(len(y_out), 3) |
---|
445 | self.assertEqual(len(dy_out), 3) |
---|
446 | |
---|
447 | def test_allowed_bins(self): |
---|
448 | x = np.asarray(np.asarray([0,1,2,3])) |
---|
449 | y = np.asarray(np.asarray([1,1,1,1])) |
---|
450 | dy = np.asarray(np.asarray([1,1,1,1])) |
---|
451 | g = invariant.Guinier() |
---|
452 | data = Data1D(x=x, y=y, dy=dy) |
---|
453 | self.assertEqual(g.get_allowed_bins(data), [False, True, True, True]) |
---|
454 | |
---|
455 | data = Data1D(x=y, y=x, dy=dy) |
---|
456 | self.assertEqual(g.get_allowed_bins(data), [False, True, True, True]) |
---|
457 | |
---|
458 | data = Data1D(x=dy, y=y, dy=x) |
---|
459 | self.assertEqual(g.get_allowed_bins(data), [False, True, True, True]) |
---|
460 | |
---|
461 | class TestDataExtraLow(unittest.TestCase): |
---|
462 | """ |
---|
463 | Generate a Guinier distribution and verify that the extrapolation |
---|
464 | produce the correct ditribution. Tested if the data generated by the |
---|
465 | invariant calculator is correct |
---|
466 | """ |
---|
467 | |
---|
468 | def setUp(self): |
---|
469 | """ |
---|
470 | Generate a Guinier distribution. After extrapolating, we will |
---|
471 | verify that we obtain the scale and rg parameters |
---|
472 | """ |
---|
473 | self.scale = 1.5 |
---|
474 | self.rg = 30.0 |
---|
475 | x = np.arange(0.0001, 0.1, 0.0001) |
---|
476 | y = np.asarray([self.scale * math.exp( -(q*self.rg)**2 / 3.0 ) for q in x]) |
---|
477 | dy = y*.1 |
---|
478 | self.data = Data1D(x=x, y=y, dy=dy) |
---|
479 | |
---|
480 | def test_low_q(self): |
---|
481 | """ |
---|
482 | Invariant with low-Q extrapolation with no slit smear |
---|
483 | """ |
---|
484 | # Create invariant object. Background and scale left as defaults. |
---|
485 | inv = invariant.InvariantCalculator(data=self.data) |
---|
486 | # Set the extrapolation parameters for the low-Q range |
---|
487 | inv.set_extrapolation(range='low', npts=10, function='guinier') |
---|
488 | |
---|
489 | self.assertEqual(inv._low_extrapolation_npts, 10) |
---|
490 | self.assertEqual(inv._low_extrapolation_function.__class__, invariant.Guinier) |
---|
491 | |
---|
492 | # Data boundaries for fiiting |
---|
493 | qmin = inv._data.x[0] |
---|
494 | qmax = inv._data.x[inv._low_extrapolation_npts - 1] |
---|
495 | |
---|
496 | # Extrapolate the low-Q data |
---|
497 | inv._fit(model=inv._low_extrapolation_function, |
---|
498 | qmin=qmin, |
---|
499 | qmax=qmax, |
---|
500 | power=inv._low_extrapolation_power) |
---|
501 | self.assertAlmostEqual(self.scale, inv._low_extrapolation_function.scale, 6) |
---|
502 | self.assertAlmostEqual(self.rg, inv._low_extrapolation_function.radius, 6) |
---|
503 | |
---|
504 | qstar = inv.get_qstar(extrapolation='low') |
---|
505 | test_y = inv._low_extrapolation_function.evaluate_model(x=self.data.x) |
---|
506 | for i in range(len(self.data.x)): |
---|
507 | value = math.fabs(test_y[i]-self.data.y[i])/self.data.y[i] |
---|
508 | self.assert_(value < 0.001) |
---|
509 | |
---|
510 | class TestDataExtraLowSlitGuinier(unittest.TestCase): |
---|
511 | """ |
---|
512 | for a smear data, test that the fitting go through |
---|
513 | real data for atleast the 2 first points |
---|
514 | """ |
---|
515 | |
---|
516 | def setUp(self): |
---|
517 | """ |
---|
518 | Generate a Guinier distribution. After extrapolating, we will |
---|
519 | verify that we obtain the scale and rg parameters |
---|
520 | """ |
---|
521 | self.scale = 1.5 |
---|
522 | self.rg = 30.0 |
---|
523 | x = np.arange(0.0001, 0.1, 0.0001) |
---|
524 | y = np.asarray([self.scale * math.exp( -(q*self.rg)**2 / 3.0 ) for q in x]) |
---|
525 | dy = y*.1 |
---|
526 | self.data = Data1D(x=x, y=y, dy=dy) |
---|
527 | self.npts = len(x)-10 |
---|
528 | |
---|
529 | def test_low_q(self): |
---|
530 | """ |
---|
531 | Invariant with low-Q extrapolation with slit smear |
---|
532 | """ |
---|
533 | # Create invariant object. Background and scale left as defaults. |
---|
534 | inv = invariant.InvariantCalculator(data=self.data) |
---|
535 | # Set the extrapolation parameters for the low-Q range |
---|
536 | inv.set_extrapolation(range='low', npts=self.npts, function='guinier') |
---|
537 | |
---|
538 | self.assertEqual(inv._low_extrapolation_npts, self.npts) |
---|
539 | self.assertEqual(inv._low_extrapolation_function.__class__, invariant.Guinier) |
---|
540 | |
---|
541 | # Data boundaries for fiiting |
---|
542 | qmin = inv._data.x[0] |
---|
543 | qmax = inv._data.x[inv._low_extrapolation_npts - 1] |
---|
544 | |
---|
545 | # Extrapolate the low-Q data |
---|
546 | inv._fit(model=inv._low_extrapolation_function, |
---|
547 | qmin=qmin, |
---|
548 | qmax=qmax, |
---|
549 | power=inv._low_extrapolation_power) |
---|
550 | |
---|
551 | |
---|
552 | qstar = inv.get_qstar(extrapolation='low') |
---|
553 | |
---|
554 | test_y = inv._low_extrapolation_function.evaluate_model(x=self.data.x[:inv._low_extrapolation_npts]) |
---|
555 | self.assert_(len(test_y) == len(self.data.y[:inv._low_extrapolation_npts])) |
---|
556 | |
---|
557 | for i in range(inv._low_extrapolation_npts): |
---|
558 | value = math.fabs(test_y[i]-self.data.y[i])/self.data.y[i] |
---|
559 | self.assert_(value < 0.001) |
---|
560 | |
---|
561 | def test_low_data(self): |
---|
562 | """ |
---|
563 | Invariant with low-Q extrapolation with slit smear |
---|
564 | """ |
---|
565 | # Create invariant object. Background and scale left as defaults. |
---|
566 | inv = invariant.InvariantCalculator(data=self.data) |
---|
567 | # Set the extrapolation parameters for the low-Q range |
---|
568 | inv.set_extrapolation(range='low', npts=self.npts, function='guinier') |
---|
569 | |
---|
570 | self.assertEqual(inv._low_extrapolation_npts, self.npts) |
---|
571 | self.assertEqual(inv._low_extrapolation_function.__class__, invariant.Guinier) |
---|
572 | |
---|
573 | # Data boundaries for fiiting |
---|
574 | qmin = inv._data.x[0] |
---|
575 | qmax = inv._data.x[inv._low_extrapolation_npts - 1] |
---|
576 | |
---|
577 | # Extrapolate the low-Q data |
---|
578 | inv._fit(model=inv._low_extrapolation_function, |
---|
579 | qmin=qmin, |
---|
580 | qmax=qmax, |
---|
581 | power=inv._low_extrapolation_power) |
---|
582 | |
---|
583 | |
---|
584 | qstar = inv.get_qstar(extrapolation='low') |
---|
585 | #Compution the y 's coming out of the invariant when computing extrapolated |
---|
586 | #low data . expect the fit engine to have been already called and the guinier |
---|
587 | # to have the radius and the scale fitted |
---|
588 | data_in_range = inv.get_extra_data_low(q_start=self.data.x[0], |
---|
589 | npts = inv._low_extrapolation_npts) |
---|
590 | test_y = data_in_range.y |
---|
591 | self.assert_(len(test_y) == len(self.data.y[:inv._low_extrapolation_npts])) |
---|
592 | for i in range(inv._low_extrapolation_npts): |
---|
593 | value = math.fabs(test_y[i]-self.data.y[i])/self.data.y[i] |
---|
594 | self.assert_(value < 0.001) |
---|
595 | |
---|
596 | |
---|
597 | class TestDataExtraHighSlitPowerLaw(unittest.TestCase): |
---|
598 | """ |
---|
599 | for a smear data, test that the fitting go through |
---|
600 | real data for atleast the 2 first points |
---|
601 | """ |
---|
602 | |
---|
603 | def setUp(self): |
---|
604 | """ |
---|
605 | Generate a Guinier distribution. After extrapolating, we will |
---|
606 | verify that we obtain the scale and rg parameters |
---|
607 | """ |
---|
608 | self.scale = 1.5 |
---|
609 | self.m = 3.0 |
---|
610 | x = np.arange(0.0001, 0.1, 0.0001) |
---|
611 | y = np.asarray([self.scale * math.pow(q ,-1.0*self.m) for q in x]) |
---|
612 | dy = y*.1 |
---|
613 | self.data = Data1D(x=x, y=y, dy=dy) |
---|
614 | self.npts = 20 |
---|
615 | |
---|
616 | def test_high_q(self): |
---|
617 | """ |
---|
618 | Invariant with high-Q extrapolation with slit smear |
---|
619 | """ |
---|
620 | # Create invariant object. Background and scale left as defaults. |
---|
621 | inv = invariant.InvariantCalculator(data=self.data) |
---|
622 | # Set the extrapolation parameters for the low-Q range |
---|
623 | inv.set_extrapolation(range='high', npts=self.npts, function='power_law') |
---|
624 | |
---|
625 | self.assertEqual(inv._high_extrapolation_npts, self.npts) |
---|
626 | self.assertEqual(inv._high_extrapolation_function.__class__, invariant.PowerLaw) |
---|
627 | |
---|
628 | # Data boundaries for fiiting |
---|
629 | xlen = len(self.data.x) |
---|
630 | start = xlen - inv._high_extrapolation_npts |
---|
631 | qmin = inv._data.x[start] |
---|
632 | qmax = inv._data.x[xlen-1] |
---|
633 | |
---|
634 | # Extrapolate the high-Q data |
---|
635 | inv._fit(model=inv._high_extrapolation_function, |
---|
636 | qmin=qmin, |
---|
637 | qmax=qmax, |
---|
638 | power=inv._high_extrapolation_power) |
---|
639 | |
---|
640 | |
---|
641 | qstar = inv.get_qstar(extrapolation='high') |
---|
642 | |
---|
643 | test_y = inv._high_extrapolation_function.evaluate_model(x=self.data.x[start: ]) |
---|
644 | self.assert_(len(test_y) == len(self.data.y[start:])) |
---|
645 | |
---|
646 | for i in range(len(self.data.x[start:])): |
---|
647 | value = math.fabs(test_y[i]-self.data.y[start+i])/self.data.y[start+i] |
---|
648 | self.assert_(value < 0.001) |
---|
649 | |
---|
650 | def test_high_data(self): |
---|
651 | """ |
---|
652 | Invariant with low-Q extrapolation with slit smear |
---|
653 | """ |
---|
654 | # Create invariant object. Background and scale left as defaults. |
---|
655 | inv = invariant.InvariantCalculator(data=self.data) |
---|
656 | # Set the extrapolation parameters for the low-Q range |
---|
657 | inv.set_extrapolation(range='high', npts=self.npts, function='power_law') |
---|
658 | |
---|
659 | self.assertEqual(inv._high_extrapolation_npts, self.npts) |
---|
660 | self.assertEqual(inv._high_extrapolation_function.__class__, invariant.PowerLaw) |
---|
661 | |
---|
662 | # Data boundaries for fiiting |
---|
663 | xlen = len(self.data.x) |
---|
664 | start = xlen - inv._high_extrapolation_npts |
---|
665 | qmin = inv._data.x[start] |
---|
666 | qmax = inv._data.x[xlen-1] |
---|
667 | |
---|
668 | # Extrapolate the high-Q data |
---|
669 | inv._fit(model=inv._high_extrapolation_function, |
---|
670 | qmin=qmin, |
---|
671 | qmax=qmax, |
---|
672 | power=inv._high_extrapolation_power) |
---|
673 | |
---|
674 | qstar = inv.get_qstar(extrapolation='high') |
---|
675 | |
---|
676 | data_in_range= inv.get_extra_data_high(q_end = max(self.data.x), |
---|
677 | npts = inv._high_extrapolation_npts) |
---|
678 | test_y = data_in_range.y |
---|
679 | self.assert_(len(test_y) == len(self.data.y[start:])) |
---|
680 | temp = self.data.y[start:] |
---|
681 | |
---|
682 | for i in range(len(self.data.x[start:])): |
---|
683 | value = math.fabs(test_y[i]- temp[i])/temp[i] |
---|
684 | self.assert_(value < 0.001) |
---|