Home / Browse / psignifit / Mail / Archive

psignifit

Email Archive: psignifit-users (read-only)

2010:	_Jan	_Feb	_Mar	_Apr (1)	_May	_Jun (2)	_Jul	_Aug (5)	_Sep	_Oct (3)	_Nov	_Dec (7)
2011:	_Jan (13)	_Feb (4)	_Mar (13)	_Apr (7)	_May (11)	_Jun (18)	_Jul (12)	_Aug (13)	_Sep (2)	_Oct (7)	_Nov (7)	_Dec (1)
2012:	_Jan (4)	_Feb (3)	_Mar (15)	_Apr (7)	_May (13)	_Jun (2)	_Jul (4)	_Aug (12)	_Sep (6)	_Oct (5)	_Nov (8)	_Dec (3)
2013:	_Jan (7)	_Feb (3)	_Mar (1)	_Apr (18)	_May (4)	_Jun	_Jul	_Aug	_Sep	_Oct	_Nov	_Dec


S	M	T	W	T	F	S
			1	2	3	4

5	6	7	8	9	10	11
(1)	(1)	(2)
12	13	14	15	16	17	18
		(6)	(6)	(1)
19	20	21	22	23	24	25

26	27	28	29	30
				(1)

[Psignifit-users] Some fails checks and errors under matlab - Psignifit3

From: Andres Martin <andres.mrtn@gm...> - 2011-06-30 13:17

Attachments: Message as HTML

Hi all!

I've installed psignifit3 in a Dell xps 501 running under Ubuntu 11.04. When
I ran "sudo make test" in the command line, the output said: "73 tests
failed" (the details are below).

On the other hand, when I ran "test_psignifit" within matlab (64bit 7.10.0
R2010a), the output was a fatal error:

>> test_psignifit
testing 2afc setting
MapEstimate
??? Error: Unexpected MATLAB operator.

Error in ==> MapEstimate at 156
eval ( output );

Error in ==> test_psignifit at 9
mapest = MapEstimate ( data2afc, priors );


Any help? Thanks in advance!!!


Make test output

[ OK ]   PsiGumbelL 1st derivative value: 0.00456628
[ OK ]   PsiGumbelL 2nd derivative value: -0.0291742
[ OK ]   PsiGumbelR 1st derivative value: 0.118205
[ OK ]   PsiGumbelR 2nd derivative value: -0.102208
[ OK ]   PsiLogistic 1st derivative value: 0.104994
[ OK ]   PsiLogistic 2nd derivative value: -0.0799625
[ OK ]   Psychometric function 1st derivative w.r.t. prm 0 value: -0.0528291
[ OK ]   Psychometric function 1st derivative w.r.t. prm 1 value: 0.0704388
[ OK ]   Psychometric function 1st derivative w.r.t. prm 2 value: -0.208609
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 0 and 0 value:
0.0205253
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 0 and 1 value:
0.00785238
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 0 and 2 value:
0.110061
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 1 and 0 value:
0.00785238
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 1 and 1 value:
-0.0574291
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 1 and 2 value:
-0.146748
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 2 and 0 value:
0.110061
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 2 and 1 value:
-0.146748
[ OK ]   Psychometric function 2nd derivative w.r.t. prm 2 and 2 value: 0
[ OK ]   psi function at x=0.5 value: -1.96351
[ OK ]   psi function at x=5.5 value: 1.61109
[ OK ]   psi function at x=10.5 value: 2.303
[ OK ]   psi function at x=15.5 value: 2.70824
[ OK ]   psi function at x=20.5 value: 2.99584
[ OK ]   psi function at x=25.5 value: 3.21894
[ OK ]   psi function at x=30.5 value: 3.40124
[ OK ]   psi function at x=35.5 value: 3.55538
[ OK ]   psi function at x=40.5 value: 3.68891
[ OK ]   psi function at x=45.5 value: 3.80668
[ OK ]   psi function at x=50.5 value: 3.91204
[ OK ]   digamma function at x=0.5 value: 4.9348
[ OK ]   digamma function at x=5.5 value: 0.199342
[ OK ]   digamma function at x=10.5 value: 0.099917
[ OK ]   digamma function at x=15.5 value: 0.066642
[ OK ]   digamma function at x=20.5 value: 0.0499896
[ OK ]   digamma function at x=25.5 value: 0.0399947
[ OK ]   digamma function at x=30.5 value: 0.0333302
[ OK ]   digamma function at x=35.5 value: 0.0285695
[ OK ]   digamma function at x=40.5 value: 0.0249987
[ OK ]   digamma function at x=45.5 value: 0.0222213
[ OK ]   digamma function at x=50.5 value: 0.0199993
=============> Solutions of optimizer <============
[ OK ]   OptimizerSolution 2AFC alpha value: 3.29751
[ OK ]   OptimizerSolution 2AFC beta value: 0.959018
[ OK ]   OptimizerSolution 2AFC lambda value: 0.0191278
[ OK ]   OptimizerSolution 2AFC deviance value: 3.9847
[ OK ]   OptimizerSolution 2AFC deviance sum value: 3.9847
[ OK ]   OptimizerSolution 2AFC Rpd value: 0.155845
[ OK ]   OptimizerSolution 2AFC Rkd value: -0.319545

[FAIL]   OptimizerSolution Y/N alpha value: 3.42566 should be: 3.43942
[FAIL]   OptimizerSolution Y/N beta value: 1.02624 should be: 0.988357
[ OK ]   OptimizerSolution Y/N lambda value: 0.000437318
[ OK ]   OptimizerSolution Y/N gamma value: 0.0240525
[FAIL]   OptimizerSolution Y/N deviance value: 2.21098 should be: 2.08172
[ OK ]   OptimizerSolution Y/N deviance sum value: 2.21098
[FAIL]   OptimizerSolution Y/N Rpd value: 0.339664 should be: 0.217146
[FAIL]   OptimizerSolution Y/N Rkd value: -0.350166 should be: -0.477967

[ OK ]   Optimizer Solution gamma=lambda, alpha value: 3.32555
[ OK ]   Optimizer Solution gamma=lambda, beta value: 1.06769
[FAIL]   Optimizer Solution gamma=lambda, lambda value: 0.000327229 should
be: 0.000506492
=============> Bootstrap properties <============
[ OK ]   Acceleration constant (threshold) value: 0.00033244 < 0.018662
[FAIL]   Bias (threshold) value: -0.0677307 should be: -0.118085
[FAIL]   th(.1) value: 2.65266 should be: 2.58437
[FAIL]   th(.9) value: 3.89757 should be: 3.83427
[ OK ]   Acceleration constant (slope) value: 0.00033244
[ OK ]   Bias (slope) value: -0.0401168
[ OK ]   sl(.1) value: 0.172577
[FAIL]   sl(.9) value: 0.497512 should be: 0.545639
[FAIL]   Deviance limits value: 9.67016 should be: 8.95713
[ OK ]   Rpd( 2.5%) value: -0.532793
[FAIL]   Rpd(97.5%) value: 0.632072 should be: 0.495472
[ OK ]   Rkd( 2.5%) value: -0.922931
[ OK ]   Rkd(97.5%) value: 0.6258
[ OK ]   influential 0 good
[ OK ]   outliers 0 good
[ OK ]   influential 1 good
[ OK ]   outliers 1 good
[ OK ]   influential 2 good
[ OK ]   outliers 2 good
[ OK ]   influential 3 good
[ OK ]   outliers 3 good
[ OK ]   influential 4 good
[ OK ]   outliers 4 good
[ OK ]   influential 5 good
[ OK ]   outliers 5 good
=============> Properties of sigmoids <============
[ OK ]   Phi(0) value: 0.5
[ OK ]   invPhi(0.5) value: 0
[ OK ]   invPhi(Phi(0.3)) value: 0.3
[ OK ]   Phi(invPhi(0.3)) value: 0.3
[ OK ]   PsiLogistic->f( 0) value: 0.5
[ OK ]   PsiLogistic->f(-3) value: 0.0474259 < 0.05
[ OK ]   PsiLogistic->f( 3) value: 0.952574 > 0.95
[ OK ]   PsiLogistic( 3)-PsiLogistic(-3) value: 0.952574
[ OK ]   PsiLogistic->df(3) value: 0.0451767 < 0.05
[ OK ]   PsiLogistic->df(-3) value: 0.0451767 < 0.05
[ OK ]   PsiLogistic monotonically increasing value: 0.00664806 > 0
[ OK ]   PsiLogistic->ddf(3) value: -0.0408916 < 0
[ OK ]   PsiLogistic->ddf(-3) value: 0.0408916 > 0
[ OK ]   PsiGauss->f( 0) value: 0.5
[ OK ]   PsiGauss->f(-3) value: 0.0013499 < 0.01
[ OK ]   PsiGauss->f( 3) value: 0.99865 > 0.99
[ OK ]   PsiGauss( 3)-PsiGauss(-3) value: 0.99865
[ OK ]   PsiGaussian monotonically increasing value: 1.48672e-06 > 0
[ OK ]   PsiGauss->df(3) value: 0.00443185 < 0.01
[ OK ]   PsiGauss->df(-3) value: 0.00443185 < 0.01
[ OK ]   PsiGauss->ddf(3) value: -0.0132955 < 0
[ OK ]   PsiGauss->ddf(-3) value: 0.0132955 > 0
[ OK ]   PsiGumbelL->f(0) value: 0.632121
[ OK ]   PsiGumbelL->f(0) value: 0.048568
[ OK ]   PsiGumbelL->f(0) value: 1
[ OK ]   PsiGumbelL( 3 )-PsiGumbelL (-3 ) value: 1 > 0.951432
[ OK ]   PsiGumbelL monotonically increasing value: 5.20543e-63 > 0
[ OK ]   PsiGumbelL->df(3) value: 3.80054e-08 < 0.01
[ OK ]   PsiGumbelL->df(-3) value: 0.047369 < 0.05
[ OK ]   PsiGumbelL->ddf(3) value: -7.25354e-07 < 0
[ OK ]   PsiGumbelL->ddf(-3) value: 0.0450106 > 0
[ OK ]   PsiGumbelR->f(0) value: 0.367879
[ OK ]   PsiGumbelR->f(0) value: 1.89218e-09
[ OK ]   PsiGumbelR->f(0) value: 0.951432
[ OK ]   PsiGumbelR( -3 )-PsiGumbelR (3 ) value: 1 > 0.951432
[ OK ]   PsiGumbelR monotonically increasing value: 5.20543e-63 > 0
[ OK ]   PsiGumbelR->df(3) value: 0.047369 < 0.05
[ OK ]   PsiGumbelR->df(-3) value: 3.80054e-08 < 0.01
[ OK ]   PsiGumbelR->ddf(3) value: -0.0450106 < 0
[ OK ]   PsiGumbelR->ddf(-3) value: 7.25354e-07 > 0
[ OK ]   PsiCauchy->f(0) value: 0.5
[ OK ]   PsiCauchy->f(-z(0.1)*(-.5) ) value: 0.1
[ OK ]   PsiCauchy->f( z(0.9)*.5 ) value: 0.9
[ OK ]   PsiCauchy monotonically increasing value: 0.0122427 > 0
[ OK ]   PsiCauchy->df(3) value: 0.031831 < 0.05
[ OK ]   PsiCauchy->df(-3) value: 0.031831 < 0.05
[ OK ]   PsiCauchy->ddf(3) value: -0.0190986 < 0
[ OK ]   PsiCauchy->ddf(-3) value: 0.0190986 > 0
[ OK ]   PsiId->f(-1) value: -1
[ OK ]   PsiId->f(0) value: 0
[ OK ]   PsiId->f( 1) value: 1
[ OK ]   PsiId->df(-1) value: 1
[ OK ]   PsiId->df( 0) value: 1
[ OK ]   PsiId->df( 1) value: 1
[ OK ]   PsiId->ddf(-1) value: 0
[ OK ]   PsiId->ddf( 0) value: 0
[ OK ]   PsiId->ddf( 1) value: 0
[ OK ]   PsiId->inv(PsiId->f(-1)) value: -1
[ OK ]   PsiId->inv(PsiId->f( 0)) value: 0
[ OK ]   PsiId->inv(PsiId->f( 1)) value: 1
=============> Tests of core objects <============
[ OK ]   abCore at threshold value: 0
[ OK ]   abCore derivative stimulus value: 0.5
[ OK ]   abCore derivative 0 value: -0.5
[ OK ]   abCore derivative 1 value: -0
[ OK ]   abCore 2nd derivative 0,0 value: 0
[ OK ]   abCore 2nd derivative 0,1 value: 0.25
[ OK ]   abCore 2nd derivative 1,1 value: 0
[ OK ]   abCore inversion g(inv(2)) value: 2
[ OK ]   abCore inversion inv(g(2)) value: 2
[ OK ]   abCore inversion dinv(2,0) value: 1
[ OK ]   abCore inversion dinv(2,1) value: 2
[ OK ]   mwCore at threshold value: 0
[ OK ]   mwCore derivative stimulus value: 2.19722
[ OK ]   mwCore derivative 0 value: -2.19722
[ OK ]   mwCore derivative 1 value: -0
[ OK ]   mwCore 2nd derivative 0,0 value: 0
[ OK ]   mwCore 2nd derivative 0,1 value: 1.09861
[ OK ]   mwCore 2nd derivative 1,1 value: 0
[ OK ]   mwCore inversion g(inv(2)) value: 2
[ OK ]   mwCore inversion inv(g(2)) value: 2
[ OK ]   mwCore inversion dinv(2,0) value: 1
[ OK ]   mwCore inversion dinv(2,1) value: 0.45512
[ OK ]   mwCore (m) for PsiLogistic value: 0.5
[ OK ]   mwCore (w) for PsiLogistic value: 2
[ OK ]   mwCore (m) for PsiGauss value: 0.5
[ OK ]   mwCore (w) for PsiGauss value: 2
[ OK ]   mwCore (m) for PsiGumbelL value: 0.5
[ OK ]   mwCore (w) for PsiGumbelL value: 2
[ OK ]   mwCore (m) for PsiCauchy value: 0.5
[ OK ]   mwCore (w) for PsiCauchy value: 2
[ OK ]   mwCore (m) for PsiExponential value: 0.5
[ OK ]   mwCore (w) for PsiExponential value: 2
[ OK ]   mwCore (m) for PsiGumbelR value: 0.5
[ OK ]   mwCore (w) for PsiGumbelR value: 2
[ OK ]   linearCore at threshold value: 0
[ OK ]   linearCore derivative stimulus value: 3
[ OK ]   linearCore derivative(0) at threshold value: -0.666667
[ OK ]   linearCore derivative(1) at threshold value: 1
[ OK ]   linearCore derivative(0,0) at threshold value: 0
[ OK ]   linearCore derivative(0,0) at threshold value: 0
[ OK ]   linearCore derivative(0,0) at threshold value: 0
[ OK ]   linearCore 2nd derivative symmetry at threshold value: 0
[ OK ]   linearCore inverse threshold value: -0.666667
[ OK ]   linearCore inverse g(inv(2)) value: 2
[ OK ]   linearCore inverse inv(g(2)) value: 2
[ OK ]   linearCore inverse derivative(0) value: 0
[ OK ]   linearCore inverse derivative(1) value: -0.333333
[ OK ]   logCore at threshold value: 0
[ OK ]   linearCore derivative stimulus value: 1
[ OK ]   logCore derivative(0) at threshold value: -0.666667
[ OK ]   logCore derivative(1) at threshold value: 1
[ OK ]   logCore derivative(0,0) at threshold value: 0
[ OK ]   logCore derivative(1,0) at threshold value: 0
[ OK ]   logCore derivative(1,1) at threshold value: 0
[ OK ]   logCore 2nd derivative symmetry at threshold value: 0
[ OK ]   logCore inverse value: 0.513417
[ OK ]   logCore inversion g(inv(2)) value: 2
[ OK ]   logCore inversion inv(g(2)) value: 2
[ OK ]   logCore inversion dinv(2,0) value: 0
[ OK ]   logCore inversion dinv(2,1) value: -0.333333
[ OK ]   NakaRushton at threshold value: 0.5
[ OK ]   NakaRushton derivative stimulus value: 0.24
[ OK ]   NakaRushton inverse value: 4
[ OK ]   NakaRushton inversion inv(g(2)) value: 2
[ OK ]   NakaRushton inversion g(inv(0.5)) value: 0.5
[ OK ]   NakaRushton inversion dinv(2,0) value: 0.5
[ OK ]   NakaRushton inversion dinv(2,1) value: 0.693147
=============> MCMC <============
[ OK ]   Hybrid MCMC alpha value: 3.5093
[ OK ]   Hybrid MCMC beta value: 0.941225
[ OK ]   Hybrid MCMC lambda value: 0.0242945
[ OK ]   Metropolis Hastings alpha value: 3.13955
[ OK ]   Metropolis Hastings beta value: 1.03687
[ OK ]   Metropolis Hastings lambda value: 0.0199283
[ OK ]   Generic Metropolis MCMC alpha value: 3.2443
[ OK ]   Generic Metropolis MCMC beta value: 0.980343
[ OK ]   Generic Metropolis MCMC lambda value: 0.0276917
=============> Priors <============
[ OK ]   Flat prior at 0 value: 1
[ OK ]   Flat prior derivative at 0 value: 0
[ OK ]   Uniform prior at -0.5 value: 0
[ OK ]   Uniform prior at 0.5 value: 1
[ OK ]   Uniform prior at 1.5 value: 0
[ OK ]   Uniform prior derivative at -0.5 value: 0
[ OK ]   Uniform prior derivative at 0.5 value: 0
[ OK ]   Uniform prior derivative at 1.5 value: 0
[ OK ]   Gaussian prior at -1 value: 0.241971
[ OK ]   Gaussian prior at 0 value: 0.398942
[ OK ]   Gaussian prior at 1 value: 0.241971
[ OK ]   Gaussian prior derivative at -1 value: 0.241971
[ OK ]   Gaussian prior derivative at 0 value: -0
[ OK ]   Gaussian prior derivative at 1 value: -0.241971
[ OK ]   BetaPrior at 0 value: 0
[ OK ]   BetaPrior at 0.1 value: 1.68095
[ OK ]   BetaPrior at 0.5 value: 1.1601
[ OK ]   BetaPrior at 1.1 value: 0
[ OK ]   BetaPrior derivative at 0 value: 0
[ OK ]   BetaPrior derivative at 0.1 value: 12.1402
[ OK ]   BetaPrior derivative at 0.5 value: 5.80049
[ OK ]   BetaPrior derivative at 1.1 value: 0
[ OK ]   GammaPrior at -0.5 value: 0
[ OK ]   GammaPrior at 0.5 value: 0.12998
[ OK ]   GammaPrior at 1.0 value: 0.1556
[ OK ]   GammaPrior at 1.5 value: 0.161314
[ OK ]   GammaPrior derivative at -0.5 value: 0
[ OK ]   GammaPrior derivative at 0.5 value: 0.0866532
[ OK ]   GammaPrior derivative at 1.0 value: 0.0259333
[ OK ]   GammaPrior derivative at 1.5 value: 0
[ OK ]   nGammaPrior at 0.5 value: 0
[ OK ]   nGammaPrior at -0.5 value: 0.12998
[ OK ]   nGammaPrior at -1.0 value: 0.1556
[ OK ]   nGammaPrior at -1.5 value: 0.161314
[ OK ]   nGammaPrior derivative at 0.5 value: -0
[FAIL]   nGammaPrior derivative at -0.5 value: -0.0866532 should be:
0.0866532
[FAIL]   nGammaPrior derivative at -1.0 value: -0.0259333 should be:
0.0259333
[ OK ]   nGammaPrior derivative at -1.5 value: -0
=============> Linear algebra routines <============
[ OK ]   Inverse (0,0) value: 1.80489
[ OK ]   Inverse (1,0) value: -0.677728
[ OK ]   Inverse (2,0) value: 0.00780493
[ OK ]   Inverse (0,1) value: -0.677728
[ OK ]   Inverse (1,1) value: 1.20944
[ OK ]   Inverse (2,1) value: 0.753826
[ OK ]   Inverse (0,2) value: 0.00780493
[ OK ]   Inverse (1,2) value: 0.753826
[ OK ]   Inverse (2,2) value: 2.48652
[ OK ]   solving Ax=b, x[0] value: 1.46603
[ OK ]   solving Ax=b, x[1] value: -0.0730086
[ OK ]   solving Ax=b, x[2] value: 0.384718
[ OK ]   Cholesky (0,0) value: 0.866025
[ OK ]   Cholesky (1,0) value: 0.600444
[ OK ]   Cholesky (2,0) value: -0.184752
[ OK ]   Cholesky (0,1) value: 0
[ OK ]   Cholesky (1,1) value: 1.00969
[ OK ]   Cholesky (2,1) value: -0.306102
[ OK ]   Cholesky (0,2) value: 0
[ OK ]   Cholesky (1,2) value: 0
[ OK ]   Cholesky (2,2) value: 0.634168
[ OK ]   LU (0,0) value: 0.75
[ OK ]   LU (1,0) value: 0.693333
[ OK ]   LU (2,0) value: -0.213333
[ OK ]   LU (0,1) value: 0.52
[ OK ]   LU (1,1) value: 1.01947
[ OK ]   LU (2,1) value: -0.303165
[ OK ]   LU (0,2) value: -0.16
[ OK ]   LU (1,2) value: -0.309067
[ OK ]   LU (2,2) value: 0.402168
[ OK ]   Ax=b, b[0] value: 1
[ OK ]   Ax=b, b[1] value: 0.5
[ OK ]   Ax=b, b[2] value: 2.77556e-17
[ OK ]   M should be symmetric value: 1
[ OK ]   matrix scaling value: 1.5
[ OK ]   matrix scaling value: 1.04
[ OK ]   matrix scaling value: -0.32
[ OK ]   matrix scaling value: 1.04
[ OK ]   matrix scaling value: 2.76
[ OK ]   matrix scaling value: -0.84
[ OK ]   matrix scaling value: -0.32
[ OK ]   matrix scaling value: -0.84
[ OK ]   matrix scaling value: 1.06
[ OK ]   pivot Ax=b, x[0] value: -5.24051
[ OK ]   pivot Ax=b, x[1] value: 1.32526
[ OK ]   pivot Ax=b, x[2] value: 0.334347
=============> Testing return bug in jackknifedata <============
[ OK ]   optimizer hung value: 0
=============> Initial parameter heuristics <============
[ OK ]   PsiPsychometric->getStart() for increasing data value: 0.608 > 0
[ OK ]   PsiPsychometric->getStart() for decreasing data value: -0.202335 <
0
=============> Finding good starting values <============
[FAIL]   yes-no: Starting value for alpha value: 3.33333 should be: 3.49558
[FAIL]   yes-no: Starting value for beta value: 1.03704 should be: 0.898865
[FAIL]   yes-no: Starting value for lambda value: 0.00185185 should be:
0.00555556
[FAIL]   yes-no: Starting value for gamma value: 0.0166667 should be:
0.0444444
[ OK ]   yes-no: minimum of alpha range value: 0
[ OK ]   yes-no: maximum of alpha range value: 10
[ OK ]   yes-no: minimum of alpha range value: 0
[ OK ]   yes-no: maximum of alpha range value: 10
[FAIL]   yes-no: minimum of beta range value: 2 should be: 0.45512
[FAIL]   yes-no: maximum of beta range value: 10 should be: 2.2756
[FAIL]   yes-no: minimum of beta range value: 2 should be: 0.45512
[FAIL]   yes-no: maximum of beta range value: 10 should be: 2.2756
[FAIL]   yes-no: minimum of lambda range value: 2 should be: 0
[ OK ]   yes-no: maximum of lambda range value: 0.1
[FAIL]   yes-no: minimum of lambda range value: 2 should be: 0
[ OK ]   yes-no: maximum of lambda range value: 0.1
[FAIL]   yes-no: minimum of gamma range value: 2 should be: 0
[ OK ]   yes-no: maximum of gamma range value: 0.1
[FAIL]   yes-no: minimum of gamma range value: 2 should be: 0
[ OK ]   yes-no: maximum of gamma range value: 0.1
[FAIL]   2afc: Starting value for alpha value: 3.33333 should be: 3.29584
[FAIL]   2afc: Starting value for beta value: 0.888889 should be: 0.988751
[FAIL]   2afc: Starting value for lambda value: 0.0203704 should be:
0.0185185
[ OK ]   2afc: minimum of alpha range value: 0
[ OK ]   2afc: maximum of alpha range value: 10
[ OK ]   2afc: minimum of alpha range value: 0
[ OK ]   2afc: maximum of alpha range value: 10
[FAIL]   2afc: minimum of beta range value: 2 should be: 0.45512
[FAIL]   2afc: maximum of beta range value: 10 should be: 2.2756
[FAIL]   2afc: minimum of beta range value: 2 should be: 0.45512
[FAIL]   2afc: maximum of beta range value: 10 should be: 2.2756
[FAIL]   2afc: minimum of lambda range value: 2 should be: 0
[ OK ]   2afc: maximum of lambda range value: 0.1
[FAIL]   2afc: minimum of lambda range value: 2 should be: 0
[ OK ]   2afc: maximum of lambda range value: 0.1
[ OK ]   grid parameter 0 lower limit value: 0
[ OK ]   grid parameter 0 upper limit value: 0.5
[ OK ]   grid parameter 1 lower limit value: 0
[ OK ]   grid parameter 1 upper limit value: 0.5
[ OK ]   grid parameter 2 lower limit value: 0
[ OK ]   grid parameter 2 upper limit value: 0.05
[ OK ]   grid.empty() on nonempty grid value: 0
[ OK ]   grid.empty() on empty grid value: 1
[ OK ]   grid.dimension() on 3d grid value: 3
[ OK ]   grid.get_gridsize() on small grid value: 5
[ OK ]   grid.front()[0] value: 0
[ OK ]   grid.front()[1] value: 0.125
[ OK ]   grid.front()[2] value: 0.25
[ OK ]   grid.front()[3] value: 0.375
[ OK ]   grid.front()[4] value: 0.5
[ OK ]   shifted grid.front()[0] value: -0.125
[ OK ]   shifted grid.front()[1] value: 0
[ OK ]   shifted grid.front()[2] value: 0.125
[ OK ]   shifted grid.front()[3] value: 0.25
[ OK ]   shifted grid.front()[4] value: 0.375
[ OK ]   shrunken grid.front()[0] value: 0
[ OK ]   shrunken grid.front()[1] value: 0.0625
[ OK ]   shrunken grid.front()[2] value: 0.125
[ OK ]   shrunken grid.front()[3] value: 0.1875
[ OK ]   shrunken grid.front()[4] value: 0.25
[ OK ]   subgrid dimension value: 2
[ OK ]   linspace 0 value: 0
[ OK ]   linspace 1 value: 0.5
[ OK ]   linspace 2 value: 1
[ OK ]   Number of generated gridpoints from 2x2x2 grid value: 8
[ OK ]   gridpoint 0 first param value: 0
[ OK ]   gridpoint 0 second param value: 0
[ OK ]   gridpoint 0 third param value: 0
[ OK ]   gridpoint 1 first param value: 0
[ OK ]   gridpoint 1 second param value: 0
[ OK ]   gridpoint 1 third param value: 0.05
[ OK ]   gridpoint 2 first param value: 0
[ OK ]   gridpoint 2 second param value: 0.5
[ OK ]   gridpoint 2 third param value: 0
[ OK ]   gridpoint 3 first param value: 0
[ OK ]   gridpoint 3 second param value: 0.5
[ OK ]   gridpoint 3 third param value: 0.05
[ OK ]   gridpoint 4 first param value: 0.5
[ OK ]   gridpoint 4 second param value: 0
[ OK ]   gridpoint 4 third param value: 0
[ OK ]   gridpoint 5 first param value: 0.5
[ OK ]   gridpoint 5 second param value: 0
[ OK ]   gridpoint 5 third param value: 0.05
[ OK ]   gridpoint 6 first param value: 0.5
[ OK ]   gridpoint 6 second param value: 0.5
[ OK ]   gridpoint 6 third param value: 0
[ OK ]   gridpoint 7 first param value: 0.5
[ OK ]   gridpoint 7 second param value: 0.5
[ OK ]   gridpoint 7 third param value: 0.05
[FAIL]   Best fit on grid value: 35.4381 should be: 23.8999
[FAIL]   Second best fit on grid value: 43.346 should be: 32.0023
[FAIL]   Best fitting alpha on grid value: 0.5 should be: 0
[ OK ]   Best fitting beta on grid value: 0.5
[ OK ]   Best fitting lambda on grid value: 0.05
[FAIL]   Second best fitting alpha on grid value: 0 should be: 0.5
[ OK ]   Second best fitting beta on grid value: 0.5
[ OK ]   Second best fitting lambda on grid value: 0.05
[FAIL]   gridpoint 0 first param value: 0.25 should be: -0.25
[ OK ]   gridpoint 0 second param value: 0.25
[ OK ]   gridpoint 0 third param value: 0.025
[FAIL]   gridpoint 1 first param value: 0.25 should be: -0.25
[ OK ]   gridpoint 1 second param value: 0.25
[ OK ]   gridpoint 1 third param value: 0.075
[FAIL]   gridpoint 2 first param value: 0.25 should be: -0.25
[ OK ]   gridpoint 2 second param value: 0.75
[ OK ]   gridpoint 2 third param value: 0.025
[FAIL]   gridpoint 3 first param value: 0.25 should be: -0.25
[ OK ]   gridpoint 3 second param value: 0.75
[ OK ]   gridpoint 3 third param value: 0.075
[FAIL]   gridpoint 4 first param value: 0.75 should be: 0.25
[ OK ]   gridpoint 4 second param value: 0.25
[ OK ]   gridpoint 4 third param value: 0.025
[FAIL]   gridpoint 5 first param value: 0.75 should be: 0.25
[ OK ]   gridpoint 5 second param value: 0.25
[ OK ]   gridpoint 5 third param value: 0.075
[FAIL]   gridpoint 6 first param value: 0.75 should be: 0.25
[ OK ]   gridpoint 6 second param value: 0.75
[ OK ]   gridpoint 6 third param value: 0.025
[FAIL]   gridpoint 7 first param value: 0.75 should be: 0.25
[ OK ]   gridpoint 7 second param value: 0.75
[ OK ]   gridpoint 7 third param value: 0.075
=============> Approximate numerical integration <============
[FAIL]   Lower bound of m value: -2.22222 should be: 0
[FAIL]   Upper bound of m value: 8.88889 should be: 10
[ OK ]   Length of m grid value: 7
[FAIL]   Lower bound of w value: -0.538267 should be: 2
[FAIL]   Upper bound of w value: 8.35062 should be: 10
[ OK ]   Length of w grid value: 7
[FAIL]   Lower bound of lm value: -0.0351852 should be: 0
[FAIL]   Upper bound of lm value: 0.0759259 should be: 0.1
[ OK ]   Length of lm grid value: 7
[ OK ]   cdf_grid Gauss value: -1.28155
[ OK ]   cdf_grid Gauss value: -0.727913
[ OK ]   cdf_grid Gauss value: -0.340695
[ OK ]   cdf_grid Gauss value: 4.27646e-18
[ OK ]   cdf_grid Gauss value: 0.340695
[ OK ]   cdf_grid Gauss value: 0.727913
[ OK ]   cdf_grid Gauss value: 1.28155
[FAIL]   fit_posterior Gauss mean value: -0.00547325 should be: 0
[FAIL]   fit_posterior Gauss std value: 0.00115231 should be: 1
[FAIL]   cdf_grid Gamma value: inf should be: 1.74477
[FAIL]   cdf_grid Gamma value: -nan should be: 2.45729
[FAIL]   cdf_grid Gamma value: -nan should be: 3.06205
[FAIL]   cdf_grid Gamma value: -nan should be: 3.67206
[FAIL]   cdf_grid Gamma value: -nan should be: 4.35885
[FAIL]   cdf_grid Gamma value: -nan should be: 5.23689
[FAIL]   cdf_grid Gamma value: -nan should be: 6.68078
[FAIL]   fit_posterior Gamma shape value: 2 should be: 4
[FAIL]   fit_posterior Gamma scale value: 1.815 should be: 1
[ OK ]   cdf_grid Beta value: 0.0256173
[ OK ]   cdf_grid Beta value: 0.0437104
[ OK ]   cdf_grid Beta value: 0.0606308
[ OK ]   cdf_grid Beta value: 0.0786438
[ OK ]   cdf_grid Beta value: 0.0996593
[ OK ]   cdf_grid Beta value: 0.127175
[ OK ]   cdf_grid Beta value: 0.172935
[FAIL]   fit_posterior Beta alpha value: 1.99882 should be: 2
[FAIL]   fit_posterior Beta beta value: 19.9891 should be: 20
[FAIL]   Posterior for m -- mu value: 3.39015 should be: 3.28
[FAIL]   Posterior for m -- sg value: 0.451554 should be: 0.436
[FAIL]   Posterior for w -- k value: 10.4441 should be: 12.732
[FAIL]   Posterior for w -- th value: 0.440721 should be: 0.362
[FAIL]   Posterior for lm -- al value: 2.54935 should be: 3.17
[FAIL]   Posterior for lm -- bt value: 106.59 should be: 111.26
H = 0.963892
[ OK ]   Sampled and fitted posterior mean for m value: 3.21063
[FAIL]   Sampled and fitted posterior mean for w value: 4.58142 should be:
4.60294
[ OK ]   Sampled and fitted posterior mean for lm value: 0.0270066
[ OK ]   number of duplicates in SIR value: 0.04 < 0.1
[ OK ]   Rkd is set value: 1
[ OK ]   Rpd is set value: 1
[ OK ]   logratio is set value: 1
[ OK ]   ppRkd is set value: 1
[ OK ]   ppRpd is set value: 1
[ OK ]   ppData is set value: 6
[ OK ]   Deviance is set value: 1
[ OK ]   ppDeviance is set value: 1

==> 73 tests failed

-- 
Andres

[Psignifit-users] using psignifit with matlab 7.8 on mac os

From: Churchland, Anne <churchland@cs...> - 2011-06-14 15:32

Hi All,

This is my first post to this group- I have got psignifit 3.0 to work with matlab but am finding things to be very slow compared with the psignifit 2. The line that is slowing things down is in BootStrapInference.m;  it is the line that calls "psignifit bootstrap". The funny thing is that the slowness doesn't seem to have to do with the number of samples. It takes 25 seconds to do the bootstrap when I have 20 samples, and only 32 seconds when I have 2000 samples! So there is some strange thing that takes 25 seconds that I can't seem to pin down. Does anyone else have this problem? I am using python 2.6.6.

Thanks!
Anne

Anne Churchland
Assistant Professor
Cold Spring Harbor Laboratories
churchland@...>
mobile: 206 853 7536
http://churchlandlab.cshl.edu

Re: [Psignifit-users] using psignifit with matlab 7.8 on mac os

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-14 16:06

Attachments: Message as HTML

Dear Anne,

this sounds like very strange behavior. Psignifit 3 is definitely slower
than psignifit 2. We try to improve on that. However, 32 seconds for
2000 samples is more than a magnitude more than I would expect. Would
you mind posting a little code example that helps us reproduce your
result? That would be very nice.

Thanks a lot for trying psignifit,

Ingo

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ingo Fründ, Dr. rer. nat.
Modelling of Cognitive Processes
TU-Berlin
Sekr 6-4
Franklinstrasse 28/29
10587 Berlin

phone: (+49|0)30-314-78962
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Re: [Psignifit-users] using psignifit with matlab 7.8 on mac os

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-15 09:28

Attachments: Message as HTML

Dear Johahn, dear Anne,

would any of you mind sending me the code you use along with an example
dataset so that I can try to do exactly the same thing you did and see
wether it takes that long for me too?

Thank you very much,

Ingo

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ingo Fründ, Dr. rer. nat.
Modelling of Cognitive Processes
TU-Berlin
Sekr 6-4
Franklinstrasse 28/29
10587 Berlin

phone: (+49|0)30-314-78962
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Re: [Psignifit-users] using psignifit with matlab 7.8 on mac os

From: Johahn Leung <johahn.leung@gm...> - 2011-06-15 01:26

Attachments: Message as HTML

Hi Ingo,

Just to further this discussion, 20-50secs is around the time it takes for a 2000 bootstrap running on my machines here too, averaging around
40secs. This is using the Matlab interface.

Running on Mac OSX or Windows didn't seem to affect the speed too much. Not having run psignifit 2, i did not know this was on the slow side.
In the case of the Mac, I compiled the command line using GCC.
With Windows, I just used the downloaded version.


We have tried fitting 40 trial staircases (fast) to Quest (slow). 

cheers

/j


___________________________
Johahn Leung
Auditory Neuroscience Laboratory
Department of Physiology
University of Sydney
T +612 93517615
F +612 93518259


On Wednesday, 15 June 2011 at 2:06 AM, Ingo Fründ wrote:

> Dear Anne,
> 
> this sounds like very strange behavior. Psignifit 3 is definitely slower
> than psignifit 2. We try to improve on that. However, 32 seconds for
> 2000 samples is more than a magnitude more than I would expect. Would
> you mind posting a little code example that helps us reproduce your
> result? That would be very nice.
> 
> Thanks a lot for trying psignifit,
> 
> Ingo
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Ingo Fründ, Dr. rer. nat.
> Modelling of Cognitive Processes
> TU-Berlin
> Sekr 6-4
> Franklinstrasse 28/29
> 10587 Berlin
> 
> phone: (+49|0)30-314-78962
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> ------------------------------------------------------------------------------
> EditLive Enterprise is the world's most technically advanced content
> authoring tool. Experience the power of Track Changes, Inline Image
> Editing and ensure content is compliant with Accessibility Checking.
> http://p.sf.net/sfu/ephox-dev2dev
> _______________________________________________
> Psignifit-users mailing list
> Psignifit-users@... (mailto:Psignifit-users@...)
> https://lists.sourceforge.net/lists/listinfo/psignifit-users

Re: [Psignifit-users] using psignifit with matlab 7.8 on mac os

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-16 14:47

Attachments: Message as HTML

Dear Anne,

in the meantime, I have reanalyzed your data in both matlab and python.
You are right. It takes about 30 seconds to fit a psychometric function
(including bootstrap) from within matlab. It takes about 4 seconds to do
the same job in python. Although 30 seconds is a very large difference,
we knew that the call from python was faster. We were not aware however,
that the difference was as big as it is. There might be two reasons for
this. First, the matlab interface is less current than the python
interface. We really came to like the fast development cycles of python.
So we usually have prototyes in python and transfer the parts that work
well to the matlab interface. This also means that the matlab interface
typically lags behind the python interface. Second, there is a lot of
ascii code that is handed around between matlab and the psignifit
command line interface. This might slow down processing even more.
However, it is not clear where a difference of 26 seconds could come
from.

Thank you also for providing installation instructions for psignifit 3
on your website. I think it is a very good idea to have such a detailed
step by step installation guide. It might be even better to have a copy
of these instructions on the psignifit website, too. What do you think?

I feel that you somehow got trapped by the rather complicated
installation instructions: You do a lot of steps that are not necessary.
We recently had a programming sprint during which the installation
instructions were restructured. It seems like some things have come out
of order there. My apologies for that. In your instructions, you can
omit ALL references to python. No need for numpy, scipy, or matplotlib.
You don't need python at all.  Psignifit is essentially written in C++.
Python is just one way to interactively use the C++ library. The other
way is the command line interface which is entirely written in C++. This
second way is what matlab uses. So what you really need is the
commandline interface and the mpsignifit folder somewhere in your matlab
path.  Typically, the command line interface should be moved to the
central folder you specified in CLI_INSTALL by just typing

make cli-install

Most probably, you only have to perform step 8 in your instructions if
you did *not* change the CLI_INSTALL variable in step 7. A good way to
check this is by typing

echo $PATH

and seeing whether the path you set via CLI_INSTALL is in the list. If
you now open a *new* terminal, you should be able to call

psignifit-mcmc -h

from anywhere on your computer.

In addition, if I remember correctly, the main problem with xcode 4 was
that you had to pay for it whereas xcode 3 comes for free. In fact, I
don't think it's a problem with the gcc version. I myself use gcc-4.4.5
to compile psignifit locally.

Best,

Ingo

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ingo Fründ, Dr. rer. nat.
Modelling of Cognitive Processes
TU-Berlin
Sekr 6-4
Franklinstrasse 28/29
10587 Berlin

phone: (+49|0)30-314-78962
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[Psignifit-users] Installation on Windows 7 64bit, Matlab 2010a

From: Ashwani Jha <a.jha@io...> - 2011-06-14 15:43

Dear Psignifit team,

I am trying to install Psignifit 3 on Windows 7 64bit, as a Matlab
2010a toolbox.
I have downloaded and extracted 'psignifit3.0_beta_07-06-2011', and
placed 'mpsignifit' in my Matlab path.

I notice that on your website you suggest also to download the command
line installer 'psignifit_cli_3_beta_installer.exe'. However the link
on the website doesn't direct me to this. Where can I get this from /
how would I install it?

best wishes,
Ashwani Jha

Re: [Psignifit-users] Installation on Windows 7 64bit, Matlab 2010a

From: Valentin Haenel <valentin.haenel@gm...> - 2011-06-15 09:42

Daer Ashwani Jha,

* Ashwani Jha <a.jha@...> [110614]:
> I am trying to install Psignifit 3 on Windows 7 64bit, as a Matlab
> 2010a toolbox.
> I have downloaded and extracted 'psignifit3.0_beta_07-06-2011', and
> placed 'mpsignifit' in my Matlab path.
> 
> I notice that on your website you suggest also to download the command
> line installer 'psignifit_cli_3_beta_installer.exe'. However the link
> on the website doesn't direct me to this. Where can I get this from /
> how would I install it?

Here is a link to the most recent version.

http://sourceforge.net/projects/psignifit/files/psignifit3.0_beta_07-06-2011/

This is windows installer and can be installed as usual.

Also, it sounds like you detected a broken link on the website. Please
write back and tell us where it is.

thanks for trying Psignifit-3.x

V-

Re: [Psignifit-users] Installation on Windows 7 64bit, Matlab 2010a

From: Ashwani Jha <a.jha@io...> - 2011-06-15 21:05

Dear Valentin,

Thanks, it works great now!
The link I tried was:

http://psignifit.sourceforge.net/INSTALL_WINDOWS.html
'Download the file psignifit_cli_3_beta_installer.exe from sourceforge
and run it.'

best wishes,
Ashwani Jha

On 15 June 2011 10:42, Valentin Haenel <valentin.haenel@...> wrote:
> Daer Ashwani Jha,
>
> * Ashwani Jha <a.jha@...> [110614]:
>> I am trying to install Psignifit 3 on Windows 7 64bit, as a Matlab
>> 2010a toolbox.
>> I have downloaded and extracted 'psignifit3.0_beta_07-06-2011', and
>> placed 'mpsignifit' in my Matlab path.
>>
>> I notice that on your website you suggest also to download the command
>> line installer 'psignifit_cli_3_beta_installer.exe'. However the link
>> on the website doesn't direct me to this. Where can I get this from /
>> how would I install it?
>
> Here is a link to the most recent version.
>
> http://sourceforge.net/projects/psignifit/files/psignifit3.0_beta_07-06-2011/
>
> This is windows installer and can be installed as usual.
>
> Also, it sounds like you detected a broken link on the website. Please
> write back and tell us where it is.
>
> thanks for trying Psignifit-3.x
>
> V-
>
>

[Psignifit-users] some help please? and a possible bug.

From: Jim Gwilliam <jim.gwilliam@jh...> - 2011-06-14 03:49

Attachments: Message as HTML

Hi,

I've spent most the day getting the psignifit 3.0 to work wiht my matlab.
 Installing the matlab path is easy, but getting the command line installed
was a bit trickier.  It turned out that matlab was trying to use some of
it's own libstd++ libraries instead of the newer linux libraries.  Once I
removed those matlab libraries to an "old" folder, it worked better.

I have two questions I hope I can receive a timely response to;

1) The first is regarding the new "Baysian Inference."  I have 5 stimulus
levels and 16 subjects.  Is it better that I use the Bayesian model instead
of the contrained ML and Bootstrap?  Also, more importantly, when I run the
"test_psignifit" in matlab, the MapEstimate, BootstrapInference,
BayesInference:explicite stepwidths all run fine (and quickly), but then
when it gets to BayesInference: pilot sample, it just says "busy" in matlab.
 My CPU shows 100% and I've left it for up to 10 minutes with no return (and
I have a powerful quad-core machine).  Any reason why it should hang like
that?

2) My second question is this:  I did a previous analysis using the old
psignifit toolbox (on a different computer).  I've been unable to duplicate
and get a working mex file on my 64 bit linux machine, so I'm using the 3.0
version now.  I have gotten that working, but I obtain different results
using 3.0 than I did with the 2.x.x version.  Most likely, I'm just doing
something different and I don't realize it.

*Previously, I used for 2.x.x:*

pfit( data, 'n_intervals', 1 );
PSE = findthreshold('logistic', data.params.est, 0.5);
JND_u = findthreshold('logistic', data.params.est, 0.25);
JND_l = findthreshold('logistic', data.params.est, 0.75);

The figures produced by pfit in 2.x.x showed the deviance and correlation
and changed text color so it was easy to tell when one of them failed.  The
3.0 doesn't see to do this, so I'm not sure what the result is when the data
plots.

Also, I'm wondering what parameters I would need to pass in 3.0 psignifit
matlab to equal the results I obtained in 2.x.x above.

*Currently, For 3.0, I'm doing:*

boots = BootstrapInference( data, priors, 'nafc', 1, 'sigmoid', 'logistic',
'core', 'ab', 'cuts', [0.25 0.5 0.75])
GoodnessOfFit( boots );

but I am not getting the same results.  I have a few questions about the
options:
a) Is conf 'v1.0' in 2.x.x default in 3.0?  How do I change conf in the 3.0
version to match?
b) I don't remember specifically setting priors in 2.x.x.  How do I match
this in 3.0?
c) is 'gamma=lambda' option default in 2.x.x?


Finally, How do I know if the Goodness of fit has passed in 3.0?  It was
clear by text color change in 2.x.x.

I'd appreciate help trying to make 3.0 match the results I had previously
obtained in 2.x.x.

Thanks,
Jim




-- 
*Jim Gwilliam*
PhD Candidate
Johns Hopkins University School of Medicine
Dept. of Biomedical Engineering
*Laboratory for Computational Sensing and
Robotics<https://haptics.lcsr.jhu.edu/Jim_Gwilliam>
137 Hackerman Hall (formerly CSEB)*
*3400 N. Charles St., Baltimore, MD 21218*
*Calendar <http://bit.ly/gwilliam_calendar>*
Office: 410-516-4184
Mobile: 801-999-8546

*
*

Re: [Psignifit-users] some help please? and a possible bug.

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-14 16:00

Attachments: Message as HTML

Hi Jim,

I'm sorry to hear that you had problems to get psignifit3 working with a
recent matlab version. With our test setup, it worked more or less out
of the box. What operatings system/ linux distribution are you using? On
what architecture? Which version of psignifit are you using? This might
help us improve the installation process and make sure psignifit takes
the "correct" libraries in the first place.

Now concerning your questions:
>    1) The first is regarding the new "Baysian Inference."  I have 5 stimulus
>    levels and 16 subjects.  Is it better that I use the Bayesian model
>    instead of the contrained ML and Bootstrap?
There are reasons to use Bayes and reasons to use Bootstrap. We are
still investigating which framework provides more "honest" confidence
regions. A first result that might be related to that question is that
Bayesian Inference seems to be more robust with respect to
nonstationarities (such as learning, fluctuating attention, changing
strategies and the like). This has recently been published in the
journal of vision if you need more details about it:
doi: 10.1167/11.6.16 Besides that, we have some (preliminary) evidence
that Bayesian inference at least does not perform worse than the
constrained ML + Bootstrap approach. However, this is (i) preliminary
and does (ii) depend on the MCMC procedure converging. We have an
alternative way of doing Bayesian Inference implemented but that is not
yet available from matlab and has not been tested in large scale
simulations. So we can't really guarantee for that. The main advantage
of the bootstrap procedure is that you don't have to worry about
converging Markov Chains. Unfortunately, bootstrap confidence intervals
are a bit biased. The bias can be reduced by performing a sensitivity
analysis as described in the original paper by Wichmann & Hill (2001).
Unfortunately, sensitivity analysis is not implemented in matlab either
(but is available from python). Sensitivity analysis is one of our main
todo points for the matlab interface. But right now that won't help you
much. Sorry.

>    Any reason why it should hang like that?
The hanging issue is not new. We had a fix for something that sounded
very similar a while ago. Do you use the most recent version? Why had a
couple of bugfixes in new downloads recently.

>    2) My second question is this:  I did a previous analysis using the old
>    psignifit toolbox (on a different computer).  I've been unable to
>    duplicate and get a working mex file on my 64 bit linux machine, so I'm
>    using the 3.0 version now.  I have gotten that working, but I obtain
>    different results using 3.0 than I did with the 2.x.x version.  Most
>    likely, I'm just doing something different and I don't realize it.
Yes, you probably do something different. If I remember correctly,
psignifit 2.x.x does the above sensitivity analysis by default. Also,
psignifit uses random numbers to generate its confidence regions. So
there will always be some random fluctations in psignifit's output.

>    The figures produced by pfit in 2.x.x showed the deviance and correlation
>    and changed text color so it was easy to tell when one of them failed.
>     The 3.0 doesn't see to do this, so I'm not sure what the result is when
>    the data plots.
Psignifit 3 is still in a development status. That means (amoung other
things) that we don't show all possible plots by default. You can
however request the plots by calling the GoodnessOfFit function. You are
right, the matlab version you are using does not yet perform the color
change. Most people actually didn't notice the color change or did not
know how to interpret it. In the python version we therefore decided to
write an explicite warning on top of the plots that tells you that
something went wrong (and a guess what might have gone wrong). That's
not in the matlab interface though. But you can easily see that by
yourself. It is described in more detail here
http://psignifit.sourceforge.net/TUTORIAL_BOOTSTRAP.html#goodness-of-fit-assessment

>    Also, I'm wondering what parameters I would need to pass in 3.0 psignifit
>    matlab to equal the results I obtained in 2.x.x above.
>    Currently, For 3.0, I'm doing:
>    boots = BootstrapInference( data, priors, 'nafc', 1, 'sigmoid',
>    'logistic', 'core', 'ab', 'cuts', [0.25 0.5 0.75])
>    GoodnessOfFit( boots );
Your parameters seem fine for me. How far apart are you the values you
get?
>    a) Is conf 'v1.0' in 2.x.x default in 3.0?  How do I change conf in the
>    3.0 version to match?
No its not. I'm not aware of the 'v1.0' switch in neither 2.x.x nor
psignifit 3. But there are differences in the default confidence limits
in 2.x.x and in 3.0:
2.x.x: [0.023, 0.159, 0.841, 0.977]
3.0  : [0.025, 0.975]

>    b) I don't remember specifically setting priors in 2.x.x.  How do I match
>    this in 3.0?
("","","Uniform(0,0.05)","Uniform(0,0.05)")
You find that in psycho_options.m

>    c) is 'gamma=lambda' option default in 2.x.x?  
No, it's not. It is not available in 2.x.x.

Please write again, in case anything was unclear or if you psignifit
does not perform the way you expected it to perform.

Best,

Ingo

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ingo Fründ, Dr. rer. nat.
Modelling of Cognitive Processes
TU-Berlin
Sekr 6-4
Franklinstrasse 28/29
10587 Berlin

phone: (+49|0)30-314-78962
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Re: [Psignifit-users] some help please? and a possible bug.

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-15 09:16

Attachments: Message as HTML

Hi Jim,

>    but in the mean time, I cannot run the old psignfit engine and I would
>    like to get it running so I can compare a single data set using psignifit,
>    the new 3.0 version, and also the python version.  (all are giving me
>    different results).
That is not really a problem. The results of psignifit depend of random
number generation. This has the effect, that your results will also be
subject to randomness. So there should be small differences between
every call you make. However, these differences should be small. If they
are not, it typically means that the number of random numbers that were
generated was not sufficiently large.

>    I am on linux 64 bit (ubuntu).  I use matla R2009b.  I get the following
>    error when trying to use the pfit function:
>    Invalid Mex File:  psignifit.mexa64: symbol mxGetName, version v7.0 not
>    defined
>    in file libmx.so with link time reference.
>    Any ideas?
Yes. Matworks (the company that produces matlab) has changed some of the
internals of matlab a couple of years ago. As a result many people
report problems running psignifit on recent versions of matlab ("recent"
here means a couple of years). To overcome this problem, we changed a
lot in the internal organization of psignifit. One thing is that we
mainly develop psignifit on python. That means that only those things
that have proven good are eventually moved to the matlab interface. The
other thing is that the real work in the matlab version is not actually
done by matlab but by an external program that is merely called from
within matlab. So, to cut a long story short: There does not seem a
solution to your problem but other people have the same problem. So if
you happen to get psignifit 2.5.6 running on a recent matlab (e.g.
R2009a), please send an eMail to this mailing list.

Besides that, thanks a lot for trying out the "new" psignifit and
helping us improve on what we have so far.

Best,

Ingo


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ingo Fründ, Dr. rer. nat.
Modelling of Cognitive Processes
TU-Berlin
Sekr 6-4
Franklinstrasse 28/29
10587 Berlin

phone: (+49|0)30-314-78962
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Re: [Psignifit-users] some help please? and a possible bug.

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-15 09:26

Attachments: Message as HTML

Dear Jim,

>    This appears to be the answer?  Is there anything I can do to fix this?
>    While updating older MEX-files, you could find calls to unsupported
>    functions, such as mxCreateFull, mxGetName, or mxIsString. MATLAB
>    removed support for these functions in Version 7.1 (R14SP3). You cannot
>    use unsupported functions with 64-bit array dimensions. For the list of
>    unsupported functions and the recommended replacements, see *Obsolete
>    Functions No Longer Documented*.
>    Update your code to use an equivalent function, if available. For example,
>    use
>    mxCreateDoubleMatrix instead of mxCreateFull.
Yes, this appears to be the answer. As I wrote in my previous mail.
However, the psignifit version you are trying to install is psignifit
2.5.6 that was written by Jeremy Hill and Felix Wichmann. It is no
longer maintained. So if you are willing to perform all the necessary
changes to get it running on a recent matlab, you are more than welcome
to do so. Be aware though, that there is a more recent version of
psignifit (psignifit 3). This more recent version is working well, but
is slightly slower than the old version. There are however, some
differences in the way you fit psychometric functions from 2.5.6 to 3.

So if you want to run psignifit on your computer, I see two options:
1. Modify the old version to run with more recent versions of matlab.
2. Use the new version with either matlab or if you are familiar with
    python, that's even better.
I would strongly advise you against option 1 -- it is a lot of work and
probably not worth it. If you need any help with option 2, please write
to this mailing list (psignifit-users@...). You will
most probably find someone to help you (for example me).

Ingo

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ingo Fründ, Dr. rer. nat.
Modelling of Cognitive Processes
TU-Berlin
Sekr 6-4
Franklinstrasse 28/29
10587 Berlin

phone: (+49|0)30-314-78962
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[Psignifit-users] Error in Matlab interface

From: Johahn Leung <johahn.leung@gm...> - 2011-06-07 05:03

Attachments: Message as HTML

Hi,

Ran test_psignifit and there's a bug in the Bayesinference.m code

specifically,

the output of line 220:
[status,output] = system ( cmd );


the first line of OUTPUT is "Hi my name is MetropolisHastings 
results.mcdata = [ 26, 33, 44, 48, 50, 46; ..."

Obviously it crashes with the "Hi my name is MetropolisHastings"

/j 
___________________________
Johahn Leung
Auditory Neuroscience Laboratory
Department of Physiology
University of Sydney
T +612 93517615
F +612 93518259

Re: [Psignifit-users] Error in Matlab interface

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-07 09:44

Attachments: Message as HTML

Hi Johan,

I'm sorry for this rather stupid bug. "Hi my name is Metropolis
Hastings" is debugging output to make sure that the code reaches the
Metropolis Hastings object at some point. Unfortunately, Matlab combines
the standard output channel with the standard error channel such that
there is no way to have any status messages from the external
programs. The bug has been fixed in the current master branch.
I will now prepare a new download snapshot that you can expect in about
two hours.

Best,

Ingo

Re: [Psignifit-users] Error in Matlab interface

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-14 14:01

Attachments: Message as HTML

Dear Johahn,

you are right, there is not sensitivity analysis yet in the matlab code.
I added that as a todo.

With respect to the deviance, you should be able to also get deviances
for all the bootstrap samples. If you use the GoodnessOfFit plot, you
will see the deviance of the ML-estimate along with a histogram of the
deviances from the bootstrap samples. That should help you interpret the
deviance value you get.

Best,

Ingo

On Fri, Jun 10, 2011 at 04:57:31PM +1000, Johahn Leung wrote:
>    Hi Ingo,
>    Thanks, I just downloaded the newest version (07_06) and the code runs.
>    However, I note that the Matlab interface does not have a Sensitivity
>    analysis, or is that incorporated in the getCI function?
>    I am getting really small CI and have no way of telling if it's correct. 
>    Using GoodnessofFit, i can get a value of deviance, but again, need a way
>    to interpret it.
>    cheers
>    /j
>    ___________________________
>    Johahn Leung
>    Auditory Neuroscience Laboratory
>    Department of Physiology
>    University of Sydney
>    T +612 93517615
>    F +612 93518259
> 
>    On Tuesday, 7 June 2011 at 8:01 PM, Ingo FrA 1/4nd wrote:
> 
>      Dear Johahn,
> 
>      the fix has been uploaded. The other bug you report below should have
>      been fixed
>      some time ago. See commit:
> 
>      833a8e9095aebb4ed907729f9fa0d3bf245a4653
> 
>      and in particular:
>      f99cd7e169c87f2821a336714e69162c78742969
> 
>      Which version are you using? If you are experienceing these problems
>      with a recent snapshot of psignifit3 could you please write me a short
>      note? What you report should not happen. We have implemented another
>      strategy for sampling from the posterior: The ASIRInference object in
>      python tries to find a reasonable parametric approximation to the
>      posterior marginals and then uses sampling importance resampling. This
>      way, there are no issues about chain convergence anymore. However, ASIR
>      inference is not yet available from matlab. If you are interested in
>      trying that out, please drop me a note.
> 
>      Thanks a lot,
> 
>      Ingo
> 
>      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>      Ingo FrA 1/4nd, Dr. rer. nat.
>      Modelling of Cognitive Processes
>      TU-Berlin
>      Sekr 6-4
>      Franklinstrasse 28/29
>      10587 Berlin
> 
>      phone: (+49|0)30-314-78962
>      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ingo Fründ, Dr. rer. nat.
Modelling of Cognitive Processes
TU-Berlin
Sekr 6-4
Franklinstrasse 28/29
10587 Berlin

phone: (+49|0)30-314-78962
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[Psignifit-users] gcc 4.6.1 errors about ptrdiff_t in swignifit_raw.cxx - patch attached

From: Wannek, Uli <wannek@un...> - 2011-06-05 16:13

Attachments: swignifit_raw.cxx.patch

Hi,

i stumbled about an error while 
> python setup.py install

Compiling "swignifit_raw.cxx" with gcc-4.6.1 raised some errors:
  ptrdiff_t does not name a type.

Searching for that error, i found a solution in http://bugs.gentoo.org/show_bug.cgi?id=362905: 
#include <cstddef>

After adding that single line, compiling the whole psignifit suite went well.

Thanks a lot for psignifit!
Best regards,
Uli

Re: [Psignifit-users] gcc 4.6.1 errors about ptrdiff_t in swignifit_raw.cxx - patch attached

From: Ingo Fründ <ingo.fruend@tu...> - 2011-06-06 09:59

Attachments: Message as HTML

Hi Uli,

thanks for reporting this problem with psignifit3. It seems like your
problem is related to swig, though. The file swignifit_raw.cxx is
automatically generated by swig. Also, you could help us by sending your
distribution. I did not find gcc-4.6.1 on the gcc website.

Best,

Ingo