Statistical Reference Datasets (StRD) from NIST

National Institute of Standards and Technology (U.S.) provides datasets for assessment of software accuracy for a variety of statistical methods. 27 datasets are provided for nonlinear regression testing: http://www.itl.nist.gov/div898/strd/nls/nls_main.shtml. This page describes how the program was tested and presents the results.

Fityk gives more accurate results than two other programs that we also tested. This is a bit surprising, because, like most of the curve fitting programs, it uses (by default) Levenberg-Marquardt method and double precision for floating point numbers. In all the cases sum of squared residuals (Residual Sum of Squares in StRD) is exactly the same as certified values, so it is not listed below.

Misra1a

Fityk can read the data directly from Misra1a.dat, automatically skipping description, but since the file has y in the first column and x in the second, we need to swap x and y. Also, we ensure that the standard deviation of all y's is 1:

@0 < Misra1a.dat
Y=x, X=y, S=1

After swapping the axes, you may click the Zoom All button in the GUI (or type plot [] in the CLI). Since this dataset has only a few points, it is a good idea to increase the size of the points (in GUI) to make them more visible.

The certified values in the NIST files are presented in the scientific format with 11 decimal places. It is possible to set the same format in Fityk:

set numeric_format='%.10E'

Now let's prepare the model with the initial values from Start 1:

define Misra1a(b1, b2) = b1*(1-exp(-b2*x))
F = Misra1a(~500, ~0.0001)

To fit the data with a very high precision we change the stopping criteria:

=-> set lm_stop_rel_change=1e-20
=-> fit
Fitting 2 (of 2) parameters to 14 points ...
In L-M method: lambda=1e+15 > 1e+15, stopped.
levenberg_marquardt: 100 iterations, 101 evaluations, 0.01 s. of CPU time.
WSSR: 1.2455138894E-01 (-99.9988%)
=-> i errors
Standard errors:
$_1 = 2.3894212918E+02 +- 2.7070075241E+00
$_2 = 5.5015643181E-04 +- 7.2668688436E-06

If Fityk was compiled with the option to use the long double type for calculations (USE_LONG_DOUBLE set to 1 in fityk.h) we get exactly the same numbers as the certified values. With the double precision used (which is the default) the parameters and standard deviations differ at 10th or 11th decimal place. All the results shown below are obtained with Fityk 1.0.0 compiled with (only) double precision.

	Parameter	Standard Deviation
b1 certified	2.3894212918E+02	2.7070075241E+00
b1 fityk/double	2.3894212923E+02	2.7070075253E+00
b2 certified	5.5015643181E-04	7.2668688436E-06
b2 fityk/double	5.5015643168E-04	7.2668688435E-06

Chwirut2

define Chwirut2(b1, b2, b3) = exp(-b1*x)/(b2+b3*x)
F = Chwirut2(~0.1, ~0.01, ~0.02)

Again, we get exactly the certified results using the long double type:

=-> i errors
Standard errors:
$_1 = 1.6657666537E-01 +- 3.8303286810E-02
$_2 = 5.1653291286E-03 +- 6.6621605126E-04
$_3 = 1.2150007096E-02 +- 1.5304234767E-03

And the only difference when using double is in b2 on 11th place.

	Parameter	Standard Deviation
b1 certified	1.6657666537E-01	3.8303286810E-02
b1 fityk/double	1.6657666537E-01	3.8303286810E-02
b2 certified	5.1653291286E-03	6.6621605126E-04
b2 fityk/double	5.1653291285E-03	6.6621605127E-04
b3 certified	1.2150007096E-02	1.5304234767E-03
b3 fityk/double	1.2150007096E-02	1.5304234767E-03

Chwirut1

define Chwirut1(b1, b2, b3) = exp(-b1*x)/(b2+b3*x)
F = Chwirut1(~0.1, ~0.01, ~0.02)

	Parameter	Standard Deviation
b1 certified	1.9027818370E-01	2.1938557035E-02
b1 fityk/double	1.9027818358E-01	2.1938557029E-02
b2 certified	6.1314004477E-03	3.4500025051E-04
b2 fityk/double	6.1314004468E-03	3.4500025053E-04
b3 certified	1.0530908399E-02	7.9281847748E-04
b3 fityk/double	1.0530908402E-02	7.9281847747E-04

Lanczos3

define Lanczos3(b1, b2, b3, b4, b5, b6) = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x)
F = Lanczos3(~1.2, ~0.3, ~5.6, ~5.5, ~6.5, ~7.6)

	Parameter	Standard Deviation
b1 certified	8.6816414977E-02	1.7197908859E-02
b1 fityk/double	8.6816415134E-02	1.7197908832E-02
b2 certified	9.5498101505E-01	9.7041624475E-02
b2 fityk/double	9.5498101593E-01	9.7041624152E-02
b3 certified	8.4400777463E-01	4.1488663282E-02
b3 fityk/double	8.4400777500E-01	4.1488663264E-02
b4 certified	2.9515951832E+00	1.0766312506E-01
b4 fityk/double	2.9515951842E+00	1.0766312488E-01
b5 certified	1.5825685901E+00	5.8371576281E-02
b5 fityk/double	1.5825685896E+00	5.8371576235E-02
b6 certified	4.9863565084E+00	3.4436403035E-02
b6 fityk/double	4.9863565087E+00	3.4436403007E-02

Gauss1

define Gauss1(b1, b2, b3, b4, b5, b6, b7, b8) = b1*exp(-b2*x) + b3*exp(-(x-b4)^2 / b5^2) + b6*exp(-(x-b7)^2 / b8^2)
F = Gauss1(~97.0, ~0.009, ~100.0, ~65.0, ~20.0, ~70.0, ~178.0, ~16.5)

This is the first dataset that gives all 11 digits in all results the same as the NIST certified values.

	Parameter	Standard Deviation
b1 certified	9.8778210871E+01	5.7527312730E-01
b1 fityk/double	9.8778210871E+01	5.7527312730E-01
b2 certified	1.0497276517E-02	1.1406289017E-04
b2 fityk/double	1.0497276517E-02	1.1406289017E-04
b3 certified	1.0048990633E+02	5.8831775752E-01
b3 fityk/double	1.0048990633E+02	5.8831775752E-01
b4 certified	6.7481111276E+01	1.0460593412E-01
b4 fityk/double	6.7481111276E+01	1.0460593412E-01
b5 certified	2.3129773360E+01	1.7439951146E-01
b5 fityk/double	2.3129773360E+01	1.7439951146E-01
b6 certified	7.1994503004E+01	6.2622793913E-01
b6 fityk/double	7.1994503004E+01	6.2622793913E-01
b7 certified	1.7899805021E+02	1.2436988217E-01
b7 fityk/double	1.7899805021E+02	1.2436988217E-01
b8 certified	1.8389389025E+01	2.0134312832E-01
b8 fityk/double	1.8389389025E+01	2.0134312832E-01

Gauss2

define Gauss2(b1, b2, b3, b4, b5, b6, b7, b8) = b1*exp(-b2*x) + b3*exp(-(x-b4)^2 / b5^2) + b6*exp(-(x-b7)^2 / b8^2)
F = Gauss2(~97.0, ~0.009, ~100.0, ~65.0, ~20.0, ~70.0, ~178.0, ~16.5)

	Parameter	Standard Deviation
b1 certified	9.9018328406E+01	5.3748766879E-01
b1 fityk/double	9.9018328406E+01	5.3748766879E-01
b2 certified	1.0994945399E-02	1.3335306766E-04
b2 fityk/double	1.0994945399E-02	1.3335306766E-04
b3 certified	1.0188022528E+02	5.9217315772E-01
b3 fityk/double	1.0188022528E+02	5.9217315773E-01
b4 certified	1.0703095519E+02	1.5006798316E-01
b4 fityk/double	1.0703095519E+02	1.5006798316E-01
b5 certified	2.3578584029E+01	2.2695595067E-01
b5 fityk/double	2.3578584028E+01	2.2695595067E-01
b6 certified	7.2045589471E+01	6.1721965884E-01
b6 fityk/double	7.2045589471E+01	6.1721965883E-01
b7 certified	1.5327010194E+02	1.9466674341E-01
b7 fityk/double	1.5327010194E+02	1.9466674342E-01
b8 certified	1.9525972636E+01	2.6416549393E-01
b8 fityk/double	1.9525972637E+01	2.6416549394E-01

DanWood

define DanWood(b1, b2) = b1*x^b2
F = DanWood(~1, ~5)

All digits the same.

	Parameter	Standard Deviation
b1 certified	7.6886226176E-01	1.8281973860E-02
b1 fityk/double	7.6886226176E-01	1.8281973860E-02
b2 certified	3.8604055871E+00	5.1726610913E-02
b2 fityk/double	3.8604055871E+00	5.1726610913E-02

Misra1b

define Misra1b(b1, b2) = b1 * (1-(1+b2*x/2)^(-2))
F = Misra1b(~500, ~0.0001)

Again, we get exactly the certified results.

	Parameter	Standard Deviation
b1 certified	3.3799746163E+02	3.1643950207E+00
b1 fityk/double	3.3799746163E+02	3.1643950207E+00
b2 certified	3.9039091287E-04	4.2547321834E-06
b2 fityk/double	3.9039091287E-04	4.2547321834E-06

Kirby2

define Kirby2(b1, b2, b3, b4, b5) = (b1 + b2*x + b3*x^2) / (1 + b4*x + b5*x^2)
F = Kirby2(~2, ~-0.1, ~0.003, ~-0.001, ~0.00001)

	Parameter	Standard Deviation
b1 certified	1.6745063063E+00	8.7989634338E-02
b1 fityk/double	1.6745063047E+00	8.7989634326E-02
b2 certified	-1.3927397867E-01	4.1182041386E-03
b2 fityk/double	-1.3927397859E-01	4.1182041376E-03
b3 certified	2.5961181191E-03	4.1856520458E-05
b3 fityk/double	2.5961181182E-03	4.1856520445E-05
b4 certified	-1.7241811870E-03	5.8931897355E-05
b4 fityk/double	-1.7241811882E-03	5.8931897332E-05
b5 certified	2.1664802578E-05	2.0129761919E-07
b5 fityk/double	2.1664802573E-05	2.0129761914E-07

Hahn1

define Hahn1(b1, b2, b3, b4, b5, b6, b7) = (b1+b2*x+b3*x^2+b4*x^3) / (1+b5*x+b6*x^2+b7*x^3)
F = Hahn1(~10, ~-1, ~0.05, ~-0.00001, ~-0.05, ~0.001, ~-0.000001)

	Parameter	Standard Deviation
b1 certified	1.0776351733E+00	1.7070154742E-01
b1 fityk/double	1.0776351730E+00	1.7070154741E-01
b2 certified	-1.2269296921E-01	1.2000289189E-02
b2 fityk/double	-1.2269296919E-01	1.2000289189E-02
b3 certified	4.0863750610E-03	2.2508314937E-04
b3 fityk/double	4.0863750606E-03	2.2508314935E-04
b4 certified	-1.4262662514E-06	2.7578037666E-07
b4 fityk/double	-1.4262662508E-06	2.7578037665E-07
b5 certified	-5.7609940901E-03	2.4712888219E-04
b5 fityk/double	-5.7609940903E-03	2.4712888218E-04
b6 certified	2.4053735503E-04	1.0449373768E-05
b6 fityk/double	2.4053735501E-04	1.0449373767E-05
b7 certified	-1.2314450199E-07	1.3027335327E-08
b7 fityk/double	-1.2314450196E-07	1.3027335327E-08

Dataset Nelson has 2 predictors (x1 and x2), Fityk does not handle 3D data, we skip this test.

MGH17

define MGH17(b1, b2, b3, b4, b5) = b1 + b2*exp(-x*b4) + b3*exp(-x*b5)
F = MGH17(~0.5, ~1.5, ~-1, ~0.01, ~0.02)

	Parameter	Standard Deviation
b1 certified	3.7541005211E-01	2.0723153551E-03
b1 fityk/double	3.7541005211E-01	2.0723153551E-03
b2 certified	1.9358469127E+00	2.2031669222E-01
b2 fityk/double	1.9358469126E+00	2.2031669217E-01
b3 certified	-1.4646871366E+00	2.2175707739E-01
b3 fityk/double	-1.4646871365E+00	2.2175707733E-01
b4 certified	1.2867534640E-02	4.4861358114E-04
b4 fityk/double	1.2867534640E-02	4.4861358109E-04
b5 certified	2.2122699662E-02	8.9471996575E-04
b5 fityk/double	2.2122699662E-02	8.9471996569E-04

Lanczos1

define Lanczos1(b1, b2, b3, b4, b5, b6) = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x)
F = Lanczos1(~1.2, ~0.3, ~5.6, ~5.5, ~6.5, ~7.6)

	Parameter	Standard Deviation
b1 certified	9.5100000027E-02	5.3347304234E-11
b1 fityk/double	9.5100000027E-02	5.3291941815E-11
b2 certified	1.0000000001E+00	2.7473038179E-10
b2 fityk/double	1.0000000001E+00	2.7444527387E-10
b3 certified	8.6070000013E-01	1.3576062225E-10
b3 fityk/double	8.6070000013E-01	1.3561973349E-10
b4 certified	3.0000000002E+00	3.3308253069E-10
b4 fityk/double	3.0000000002E+00	3.3273686649E-10
b5 certified	1.5575999998E+00	1.8815731448E-10
b5 fityk/double	1.5575999998E+00	1.8796204982E-10
b6 certified	5.0000000001E+00	1.1057500538E-10
b6 fityk/double	5.0000000001E+00	1.1046025358E-10

Lanczos2

define Lanczos2(b1, b2, b3, b4, b5, b6) = b1*exp(-b2*x) + b3*exp(-b4*x) + b5*exp(-b6*x)
F = Lanczos2(~1.2, ~0.3, ~5.6, ~5.5, ~6.5, ~7.6)

	Parameter	Standard Deviation
b1 certified	9.6251029939E-02	6.6770575477E-04
b1 fityk/double	9.6251029939E-02	6.6770575433E-04
b2 certified	1.0057332849E+00	3.3989646176E-03
b2 fityk/double	1.0057332849E+00	3.3989646154E-03
b3 certified	8.6424689056E-01	1.7185846685E-03
b3 fityk/double	8.6424689056E-01	1.7185846674E-03
b4 certified	3.0078283915E+00	4.1707005856E-03
b4 fityk/double	3.0078283915E+00	4.1707005829E-03
b5 certified	1.5529016879E+00	2.3744381417E-03
b5 fityk/double	1.5529016879E+00	2.3744381401E-03
b6 certified	5.0028798100E+00	1.3958787284E-03
b6 fityk/double	5.0028798100E+00	1.3958787275E-03

Gauss3

define Gauss3(b1, b2, b3, b4, b5, b6, b7, b8) = b1*exp(-b2*x) + b3*exp(-(x-b4)^2 / b5^2) + b6*exp(-(x-b7)^2 / b8^2)
F = Gauss3(~97.0, ~0.009, ~100.0, ~65.0, ~20.0, ~70.0, ~178.0, ~16.5)

	Parameter	Standard Deviation
b1 certified	9.8940368970E+01	5.3005192833E-01
b1 fityk/double	9.8940368970E+01	5.3005192832E-01
b2 certified	1.0945879335E-02	1.2554058911E-04
b2 fityk/double	1.0945879335E-02	1.2554058910E-04
b3 certified	1.0069553078E+02	8.1256587317E-01
b3 fityk/double	1.0069553078E+02	8.1256587330E-01
b4 certified	1.1163619459E+02	3.5317859757E-01
b4 fityk/double	1.1163619459E+02	3.5317859756E-01
b5 certified	2.3300500029E+01	3.6584783023E-01
b5 fityk/double	2.3300500027E+01	3.6584783020E-01
b6 certified	7.3705031418E+01	1.2091239082E+00
b6 fityk/double	7.3705031424E+01	1.2091239080E+00
b7 certified	1.4776164251E+02	4.0488183351E-01
b7 fityk/double	1.4776164250E+02	4.0488183353E-01
b8 certified	1.9668221230E+01	3.7806634336E-01
b8 fityk/double	1.9668221231E+01	3.7806634338E-01

Misra1c

define Misra1c(b1, b2) = b1 * (1-(1+2*b2*x)^(-.5))
F = Misra1c(~500, ~0.0001)

	Parameter	Standard Deviation
b1 certified	6.3642725809E+02	4.6638326572E+00
b1 fityk/double	6.3642725809E+02	4.6638326572E+00
b2 certified	2.0813627256E-04	1.7728423155E-06
b2 fityk/double	2.0813627256E-04	1.7728423155E-06

Misra1d

define Misra1d(b1, b2) = b1*b2*x*((1+b2*x)^(-1))
F = Misra1d(~500, ~0.0001)

	Parameter	Standard Deviation
b1 certified	4.3736970754E+02	3.6489174345E+00
b1 fityk/double	4.3736970754E+02	3.6489174345E+00
b2 certified	3.0227324449E-04	2.9334354479E-06
b2 fityk/double	3.0227324449E-04	2.9334354479E-06

Roszman1

define Roszman1(b1, b2, b3, b4) = b1 - b2*x - atan(b3/(x-b4))/pi
F = Roszman1(~0.1, ~-0.00001, ~1000, ~-100)

	Parameter	Standard Deviation
b1 certified	2.0196866396E-01	1.9172666023E-02
b1 fityk/double	2.0196866395E-01	1.9172666023E-02
b2 certified	-6.1953516256E-06	3.2058931691E-06
b2 fityk/double	-6.1953516245E-06	3.2058931692E-06
b3 certified	1.2044556708E+03	7.4050983057E+01
b3 fityk/double	1.2044556708E+03	7.4050983056E+01
b4 certified	-1.8134269537E+02	4.9573513849E+01
b4 fityk/double	-1.8134269538E+02	4.9573513852E+01

ENSO

define Enso(b1, b2, b3, b4, b5, b6, b7, b8, b9) = b1 + b2*cos(t/12) + b3*sin(t/12) + b5*cos(t/b4) + b6*sin(t/b4) + b8*cos(t/b7) + b9*sin(t/b7) where t=2*pi*x
F = Enso(~11.0, ~3.0, ~0.5, ~40.0, ~-0.7, ~-1.3, ~25.0, ~-0.3, ~1.4)

	Parameter	Standard Deviation
b1 certified	1.0510749193E+01	1.7488832467E-01
b1 fityk/double	1.0510749192E+01	1.7488832468E-01
b2 certified	3.0762128085E+00	2.4310052139E-01
b2 fityk/double	3.0762128089E+00	2.4310052139E-01
b3 certified	5.3280138227E-01	2.4354686618E-01
b3 fityk/double	5.3280138256E-01	2.4354686618E-01
b4 certified	4.4311088700E+01	9.4408025976E-01
b4 fityk/double	4.4311088678E+01	9.4408025781E-01
b5 certified	-1.6231428586E+00	2.8078369611E-01
b5 fityk/double	-1.6231428614E+00	2.8078369522E-01
b6 certified	5.2554493756E-01	4.8073701119E-01
b6 fityk/double	5.2554493176E-01	4.8073701153E-01
b7 certified	2.6887614440E+01	4.1612939130E-01
b7 fityk/double	2.6887614416E+01	4.1612939012E-01
b8 certified	2.1232288488E-01	5.1460022911E-01
b8 fityk/double	2.1232286294E-01	5.1460022934E-01
b9 certified	1.4966870418E+00	2.5434468893E-01
b9 fityk/double	1.4966870451E+00	2.5434468712E-01

MGH09

define MGH09(b1, b2, b3, b4) = b1*(x^2+x*b2) / (x^2+x*b3+b4)
F = MGH09(~25, ~39, ~41.5, ~39)

	Parameter	Standard Deviation
b1 certified	1.9280693458E-01	1.1435312227E-02
b1 fityk/double	1.9280693477E-01	1.1435312166E-02
b2 certified	1.9128232873E-01	1.9633220911E-01
b2 fityk/double	1.9128232455E-01	1.9633220578E-01
b3 certified	1.2305650693E-01	8.0842031232E-02
b3 fityk/double	1.2305650614E-01	8.0842030241E-02
b4 certified	1.3606233068E-01	9.0025542308E-02
b4 fityk/double	1.3606232875E-01	9.0025541048E-02

Thurber

define Thurber(b1, b2, b3, b4, b5, b6, b7) = (b1 + b2*x + b3*x^2 + b4*x^3) / (1 + b5*x + b6*x^2 + b7*x^3)
F = Thurber(~1000, ~1000, ~400, ~40, ~0.7, ~0.3, ~0.03)

	Parameter	Standard Deviation
b1 certified	1.2881396800E+03	4.6647963344E+00
b1 fityk/double	1.2881396799E+03	4.6647963334E+00
b2 certified	1.4910792535E+03	3.9571156086E+01
b2 fityk/double	1.4910792548E+03	3.9571155198E+01
b3 certified	5.8323836877E+02	2.8698696102E+01
b3 fityk/double	5.8323836966E+02	2.8698695459E+01
b4 certified	7.5416644291E+01	5.5675370270E+00
b4 fityk/double	7.5416644466E+01	5.5675368991E+00
b5 certified	9.6629502864E-01	3.1333340687E-02
b5 fityk/double	9.6629502951E-01	3.1333340084E-02
b6 certified	3.9797285797E-01	1.4984928198E-02
b6 fityk/double	3.9797285837E-01	1.4984927908E-02
b7 certified	4.9727297349E-02	6.5842344623E-03
b7 fityk/double	4.9727297607E-02	6.5842343110E-03

BoxBOD

define BoxBOD(b1, b2) = b1*(1-exp(-b2*x))
F = BoxBOD(~100, ~0.75)

	Parameter	Standard Deviation
b1 certified	2.1380940889E+02	1.2354515176E+01
b1 fityk/double	2.1380940890E+02	1.2354515178E+01
b2 certified	5.4723748542E-01	1.0455993237E-01
b2 fityk/double	5.4723748534E-01	1.0455993235E-01

Rat42

define Rat42(b1, b2, b3) = b1 / (1+exp(b2-b3*x))
F = Rat42(~100, ~1, ~0.1)

	Parameter	Standard Deviation
b1 certified	7.2462237576E+01	1.7340283401E+00
b1 fityk/double	7.2462237576E+01	1.7340283401E+00
b2 certified	2.6180768402E+00	8.8295217536E-02
b2 fityk/double	2.6180768402E+00	8.8295217536E-02
b3 certified	6.7359200066E-02	3.4465663377E-03
b3 fityk/double	6.7359200066E-02	3.4465663377E-03

MGH10

define MGH10(b1, b2, b3) = b1 * exp(b2/(x+b3))
F = MGH10(~0.02, ~4000, ~250)

	Parameter	Standard Deviation
b1 certified	5.6096364710E-03	1.5687892471E-04
b1 fityk/double	5.6096364710E-03	1.5687892478E-04
b2 certified	6.1813463463E+03	2.3309021107E+01
b2 fityk/double	6.1813463463E+03	2.3309021118E+01
b3 certified	3.4522363462E+02	7.8486103508E-01
b3 fityk/double	3.4522363462E+02	7.8486103544E-01

Eckerle4

define Eckerle4(b1, b2, b3) = b1 * exp(b2/(x+b3))
F = Eckerle4(~1.5, ~5, ~450)

	Parameter	Standard Deviation
b1 certified	1.5543827178E+00	1.5408051163E-02
b1 fityk/double	1.5543827179E+00	1.5408051165E-02
b2 certified	4.0888321754E+00	4.6803020753E-02
b2 fityk/double	4.0888321764E+00	4.6803020765E-02
b3 certified	4.5154121844E+02	4.6800518816E-02
b3 fityk/double	4.5154121844E+02	4.6800518828E-02

Rat43

define Rat43(b1, b2, b3, b4) = b1 / ((1+exp(b2-b3*x))^(1/b4))
F = Rat43(~100, ~10, ~1, ~1)

	Parameter	Standard Deviation
b1 certified	6.9964151270E+02	1.6302297817E+01
b1 fityk/double	6.9964151266E+02	1.6302297809E+01
b2 certified	5.2771253025E+00	2.0828735829E+00
b2 fityk/double	5.2771253072E+00	2.0828735841E+00
b3 certified	7.5962938329E-01	1.9566123451E-01
b3 fityk/double	7.5962938380E-01	1.9566123468E-01
b4 certified	1.2792483859E+00	6.8761936385E-01
b4 fityk/double	1.2792483872E+00	6.8761936428E-01

Bennett5

define Bennett5(b1, b2, b3) = b1 * (b2+x)^(-1/b3)
F = Bennett5(~-2000, ~50, ~0.8)

	Parameter	Standard Deviation
b1 certified	-2.5235058043E+03	2.9715175411E+02
b1 fityk/double	-2.5235058043E+03	2.9715175183E+02
b2 certified	4.6736564644E+01	1.2448871856E+00
b2 fityk/double	4.6736564644E+01	1.2448871760E+00
b3 certified	9.3218483193E-01	2.0272299378E-02
b3 fityk/double	9.3218483193E-01	2.0272299222E-02

NIST certified data - wojdyr/fityk GitHub Wiki

Statistical Reference Datasets (StRD) from NIST

Misra1a

Chwirut2

Chwirut1

Lanczos3

Gauss1

Gauss2

DanWood

Misra1b

Kirby2

Hahn1

MGH17

Lanczos1

Lanczos2

Gauss3

Misra1c

Misra1d

Roszman1

ENSO

MGH09

Thurber

BoxBOD

Rat42

MGH10

Eckerle4

Rat43

Bennett5

⚠️ GitHub.com Fallback ⚠️

NIST certified data - wojdyr/fityk GitHub Wiki

Statistical Reference Datasets (StRD) from NIST

⚠️ **GitHub.com Fallback** ⚠️

⚠️ GitHub.com Fallback ⚠️