minkwe wrote:So then answer this simple question if you can, honestly.Why oh why are you unable to make your point by analysing the output data from MY PROGRAMS. Why is it necessary for you to rewrite the program in order to analyse it?
Quite a few reasons.
(0) I wanted to analyse the model, not the program.
(1) I wanted to run CHSH-style experiments. Alice and Bob each choosing between two specific settings.
(2) I wanted to do my own data-analyses on the data, not yours; and I wanted to do them in R.
(3) I wanted to analyse the data in R, on a MacBook Pro under OS X.9. I am not a Python wizard. I couldn't get the R package for reading Numpy data files into R to work under my version of R
(4) I wanted to repeat experiments many times. It costs an awful lot of time to save large files to disks with one program and then read them off disk with another. Much better to have them created in memory and analysed in memory, and the repetitions all done within the same program.
(5) I wanted to make my experiments reproducible and for that purpose I needed to add set/save/restore random seed functionality
(6) I noticed that some of your numerical and probabilistic procedures are rather inefficiently programmed. There are much faster and at the same time numerically much more accurate ways to simulate your model.
(7) There are very decent interfaces between R and C++, and between R some other languages, but not between R and Python. And I'm not a Python wizard.
I use R because it is designed for statistical data analysis and programming. It's perfect for my present purposes. Programs are very short, easy to write and easy to debug and easy to explain to other people. Rpubs and Rstudio are wonderful tools for communication and development and more tools are on the way e.g. RShiny is an R server so that you can put interactive R programs online.
I haven't told you here what the point was. There were quite a few interesting points and some new discoveries.
Incidentally, Hans de Raedt agreed with me that Aspect and Weihs applied the wrong inequality. If they had used the appropriate inequality they would not have got a statistically significant violation. To put it another way: they did not test local realism, they tested local realism plus some other assumptions. So if the inequality which they tested was violated, then the other assumptions could be at fault, not local realism.