Okay, so I’m now quite dismayed by my geeky-ness. I got very excited when Distance gave me:
Effort : 210.0000
# samples : 1
Width : 4.000000
# observations: 266
Model
Half-normal key, k(y) = Exp(-y**2/(2*A(1)**2))
Cosine adjustments of order(s) : 3
A( 1) bounds = (0.40000E-01 , 0.10000E+07 )
Results:
Convergence was achieved with 12 function evaluations.
Final Ln(likelihood) value = -340.41388
Akaike information criterion = 684.82776
Bayesian information criterion = 691.99475
AICc = 684.87341
Final parameter values: 2.1114910 0.34375469E-01
And my mrds run (in debug) gave me:
D(19)> getpar(model$par,ltmodel$aux$ftype,ltmodel$aux$zdim)
$key.scale
(Intercept)
0.747398
$key.shape
NULL
$adj.parm
0.03437112
The value of adj.parm and the second “Final parameter” value are the same and also if you raise the value of key.scale to the power e then you get the first “Final parameter” value.
For those who haven’t yet caught on, this means that the code I’ve been writing gives the same answer as the equivalent FORTRAN code. This is good.
Share This