Using Fit Censored for other distributions?

Question

Hi.  I'm interested in using the Fit Censored() function to fit distributions to my data.  So far it appears I can fit only to these distributions though: Weibull, LogNormal, LogLogistic, Frechet, SEV, Normal, Logistic, and LEV.

My question: is there a way to fit to JohnsonSU/SL/SB or normal multimodal using this function?  For example, this script returns an error for me:

Names Default To Here(1);
result =
Fit Censored(
	Distribution("JohnsonSU"),
	Y(
		[142, 156, 163, 198, 204, 205, 232, 239,
		240, 261, 280, 296, 323, 344]
	)
);
Show(result);

Error: Distribution 'JohnsonSU' unrecognized in access or evaluation of 'Distribution' , Distribution(/*###*/"JohnsonSU")

Or is it possible I'm overlooking some other function that can fit a distribution to my data?  While it is true I can always use the Distribution platform to fit my data to alternative distributions, the Fit Censored function appears to be much faster.  I assume it has less overhead.

Any advice is appreciated.  Thanks.

Mark_Bailey · Answer

This function is not documented very well. Here is the Help > JSL Syntax Guide entry for it.

This is the same as the Help > Scripting Index entry:

nikles · Answer

Thanks for the help. Yes, I think my data set is probably pretty large IMO (10-20k columns, and 10-100k rows. I suppose it's all relative though). When I'm trying to fit distributions to that data, every millisecond counts. I've not yet speed tested the Fit Censored method vs the Distribution platform method, but my instinct is that the former will be faster. Plus, I find the former method easier programmatically to extract the fit parameters.

nikles · Answer

Replying to my own question.

I did some more digging and found this function:  Fit Transform to Normal().

Names Default To Here(1);
datavec = [-3.7975076, 0.48221038, -1.3082712,
-1.860647, -6.9470789, -17.237024, -19.470857,
-6.1855986, 2.16525629, -30.990061];
freqvec = [1, 1, 1, 1, 1, 2, 2, 2, 2, 2];
As Table(datavec || freqvec);
Column(1) << set name("x");
Column(2) << set name("freq vec");
results =
Fit Transform To Normal(
	Distribution("glog"),
	Y(datavec),
	freq(freqvec)
);
Show(results);

I'm trying to figure out what it does though.  At first glance, it appears to fit a vector of data to Johnson SL/SB/SU or GLog distributions, returning the parameters for that respective distribution.  If so, this is exactly what I want.  Is anyone more familiar with this function though, and do I understand it correctly?

I'm confused bc the function name seems to imply it is transforming the vector into a normal distribution, but if that were the case, how would that be different from simply fitting your data to a normal distribution?  I don't fully understand when this function would be used.  If anyone out there knows, please share!