Could someone test if they can open this notebook? There is something wrong with it. The first time I try to open it, it doesn't open. The second time it does. dl.dropboxusercontent.com/u/38623/IGDocumentation.nb (3.6 MB)
@C.E. @Kuba I found the problem. It had Visible -> False set. This happened while evaluating the notebook from the command line with UsingFrontEnd (and an automatically hidden front end) ...
Finnish government is about to have an experiment on basic income, and is selecting 2000 test subjects from group of eligible candidates. They have published source code for selection method (in SAS).
The problem is that the selection method is absurdly broken, by lack of entropy. Code manages to have effective amount of entropy of around 15 bits for the selection process, when they would actually need 20+ kilobits of it, and routines for subset selection which simply aren't there in sufficient fashion when using SAS surveyselect and PRNG it uses internally.
I had to write a violent rant on it to them. I wonder if that selection routine is actually mandated by law to be exactly what they describe...
I'm not saying that anyone would be involved in active wrongdoing, but randomness of subject selection is very illusory.
@C.E. The problem in this regard is that the nuances that make the thing broken are subtle enough that they're basically opaque technobabble for the non-initiated.
I'd say their intent is noble in publishing the sampling code, but sadly instead of increasing confidence in impartiality of the solution, it rather removes trust in professionalism.
@Szabolcs Concerning this question mathematica.stackexchange.com/q/15628/78 you have edited it as fixed since V11. But I think that the following plot still shows some symptoms: LogPlot[E^(-1000 x^2), {x, -300, 300}]
If we use -100, instead of -1000 the scale is indeed logarithmic. But with the -1000, in 11.0.1 (windows), I get a linear scale.
I don't have 11.1beta installed, so I haven't checked if it is something underwork. If someone can check on the most recent "released" build, that would be great.
Also, with Frame->True, we can see a bad side effect of this: LogPlot[E^(-1000 x^2), {x, -300, 300}, Frame -> True]
(I get no ticks at all)
Using 1000.0 instead of 1000, changes the behaviour, as nothing is plotted. And this is something that doesn't seem to be correctable with WorkingPrecision.
When we are on a user's page, what does the "Last seen" information mean?
Does this information indicate that the user (profile owner) logged into the system? Or that any other user saw something from this user? Or does it indicate neither of these things, but something else?
@yode I don't know why you're concerned about positive and negative values. It's greater than and less than that matter in determining more and less complex.
@acl yeah, but how do I know that there isn't some other bug in Array, or that it's not going to have some other funny interaction with my existing code?
I've been using the combination
ConstantArray[0, {}]
(* 0 *)
in my code for some time, as a way for returning zero (i.e. a zero-dimensional array of zeroes) inside scalable code that can be made to produce empty arrays of arbitrary dimensions but also zeros. On my system (currently v11 over l...
@halirutan I think we have every right to be confused about cases like this. The problem is that the query has an operator of the form descending /* ascending /* descending. The documentation is silent on how such forms are treated and the actual interpretation has been changing across releases.
Presently, such forms are treated as pure ascending operators:
But by converting Values to be ascending, the operator takes the form descending /* ascending /* ascending which is a form that is documented to preserve the descending character of the first operator:
So, the upshot is that by converting the trailing operator to be ascending, the leading operator becomes descending. Hardly intuitive.
I think it would be more intuitive (less unintuitive?) if descending /* ascending /* descending were treated as descending /* ascending /* ascending. But that would mean yet another behavioural change. Sigh.
@WReach Yes, the crucial point I was missing is how this particular combination of operators is interpreted as a whole. Your explanation is very illustrative and I knew that you are the one who knows a great deal about this. Thank you again for taking the time.
@halirutan The PackageData chat room was closed six days after an automated RSS message, which was fourteen days since the last message by a user. In other words, automated messages do not count as activity in the sense that it will prevent the room from being frozen.