Even manual is not freely available?
Seriously in the war of Stata versus SAS versus R who uses limdep?
-
Laugh all you want but before Stata it was either Limdep or write and maximize the f**king ML function in Fortran.
Writing your own in Fortran or C will always run faster and be more flexible than any black box option. I don’t use statistical software, it is a crutch for the weak of mind.
-
^ Coding monkey? I prefer to be an economist.
Statistical software only works for canned reg y x, robust kind of stuff. Even a moderately complex Bayesian model is best written and compiled to binary. Many reasons for this: execution speed, full control over the sampling algorithm and its output, tailoring the sampler to the specific form of the model, etc.
-
^ Coding monkey? I prefer to be an economist.
Statistical software only works for canned reg y x, robust kind of stuff. Even a moderately complex Bayesian model is best written and compiled to binary. Many reasons for this: execution speed, full control over the sampling algorithm and its output, tailoring the sampler to the specific form of the model, etc.
Execution speed is only part of the story though. You should be thinking of project completion time. If the code runs 4 hours faster in binary over a canned package, but takes 4 weeks long to write, are we really better off?
-
^ Coding monkey? I prefer to be an economist.
Statistical software only works for canned reg y x, robust kind of stuff. Even a moderately complex Bayesian model is best written and compiled to binary. Many reasons for this: execution speed, full control over the sampling algorithm and its output, tailoring the sampler to the specific form of the model, etc.
Execution speed is only part of the story though. You should be thinking of project completion time. If the code runs 4 hours faster in binary over a canned package, but takes 4 weeks long to write, are we really better off?
Not only completion time we need to be concerned with reproducibility as well. who cares if you complete the project quickly if the results aren't reproducible. In this case you are just misleading yourself that you have done a good job by finishing the project quickly.