Machine learning in trading: theory, models, practice and algo-trading - page 207
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I'm telling you, you're trying hard to get away from a specific discussion.
All right, at least you acknowledged one flaw. Forgot only to admit that we have experts who are able to check, understand and find a better solution.
Alexei, wait for an answer from R. And notice how you stopped answering @Quantum's questions. He deliberately leads you neatly to a known goal.
So far on our side is Mathematica + Wolfram Alpha + Mathlab + MQL5, and on yours is the opsorced R. The code is written carelessly and isn't as well polished as you'd expect from a 20 year old project.I will answer Quantum.
You yourself see the discrepancy, that for the jamb you referred to the article and your arguments can be checked, and for other "errors" we are forced to believe the words of Quantum. And where are the references to publications? Where is the science, I see only dogmatic statements.
And so, references where density studies at extremes are done and a reference to the R package is desirable. A review of the Loader article where? Or are my questions irrelevant?
I will answer Quantum.
You yourself see the discrepancy, that for the jamb you referred to the article and your arguments can be checked, and for the other "errors" we are forced to believe the words of Quantum. And where are the references to publications? Where is the science, I see only dogmatic statements.
And so, references where density studies at extremes are done and a reference to the R package is desirable. A review of the Loader article where? Or are my questions irrelevant?
Please answer.
And don't pretend that there are only words (with justifications, by the way), and there are no confirmations in Mathematica + Wolfram Alpha + Mathlab. You already understand everything.
You don't bother to look at those very materials and cite them, unlike @Quantum.
And you even make references to Excel and Python so as not to give a clear example.
So far you are the only one practicing wit, too.
Don't forget to bring an answer from R, if you get it, of course.
No problem, Renat. You don't believe me, look:
https://en.wikipedia.org/wiki/Gamma_distribution a support for (0,inf)
http://mathworld.wolfram.com/GammaDistribution.html helpd for [0,inf]
http://www.math.uah.edu/stat/special/Gamma.html support for [0,inf]
R http://stat.ethz.ch/R-manual/R-patched/library/stats/html/GammaDist.html helpport for [0,inf]
http://pj.freefaculty.org/guides/stat/Distributions/DistributionWriteups/Gamma/Gamma-02.pdf helpport for [0,inf]
http://www.csie.ntu.edu.tw/~sdlin/download/Probability%20&%20Statistics.pdf helpport for [0,inf]
etc.
And to your hypocritical comment that you have Wolfram, Matlab, etc. on your side, I will again bother to reply that they are not the truth on this issue. That's what they agreed on. Other versions are no less correct. If this is unclear to you, then don't bother loading me with your comments at all.
Would you be so kind as to reply?
And don't pretend that there are only words (with justifications, by the way), and no confirmation in Mathematica + Wolfram Alpha + Mathlab. You already understand everything.
I looked at Quantum's posts and found the ones I have yet to respond to.
And where are the links to publications that substantiate your statements, not finger pointing, in the article I don't see.
How would the developers of R explain their results:
dgamma(0,0.5,1)=inf
pgamma(0,0.5,1)=0
if they have point 0 included (as seen in the definition), gives infinite density at point x=0, and then when integrating in pgamma(x,0.5,1) infinity is considered zero, as if it did not exist.
In my opinion, it is normal that the integral on the left at zero will be zero. And it's okay for the density to be infinite.
Read this:
Source
dgamma is computed via the Poisson density, using code contributed by Catherine Loader (see dbinom).
----
Source
For dbinom a saddle-point expansion is used: see
Catherine Loader (2000). Fast and Accurate Computation of Binomial Probabilities; available from http://www.herine.net/stat/software/dbinom.html.
Give a reasonable answer that there is a flaw in the proposed algorithm.
And please name the source of the algorithm you used for gamma distribution density in MQL.
And to your hypocritical comment that you have Wolfram, Matlab, etc. on your side, I will again bother to answer that they are not the truth in this matter. That's what they agreed on. Other versions are no less correct. If this is unclear to you, then don't bother loading me with your comments at all.
That's just great. Already acknowledged mats are not true. And you believe in the code of an Open Source solution R, moreover you haven't seen this code yourself unlike us.
You even tried with unrepresented(why?) code in Excel and Python to convince your position. So in this message you ignored "give me the code, what you have counted in Excel and Python".
So collectively you have a very weak position and you know it.
Read your posts, please.
And be kind not only to give links (read this one and talk to yourself!), but clear statements from the given links with proof and rebuttal of @Quantum's claims. Don't walk away from clear statements.
I mean, we know why, instead of clear positions, you go around trying to get out of the discussion in a "ah, that's just your blather, they agreed, what's there to discuss" mode.
That's just great. Already recognized matrices are not the truth. And you believe in the code of R's outsourced solution, and you haven't seen this code yourself, unlike us.
You even tried with unrepresented (why?) code in Excel and Python to convince your position. So in this message you ignored "give me the code, what you have counted in Excel and Python".
So in the aggregate you have a very weak position and you know it.
Read your posts, please.
And be kind not only to give links (read this one and talk to yourself!), but clear statements from the given links with proof and rebuttal of @Quantum's claims. Don't walk away from clear statements.
After all, we know why, instead of clear positions, you go around trying to get out of the discussion in the mode of "oh, that's just your nonsense, they agreed, what is there to discuss".
This is already lazy. Go to the link and find the line where support is listed. I will not comment on it, sorry.
About Excel - why I named it.
According to the logic voiced by the man Quantum, who somehow implements the gamma distribution function in your software, if the integral on the left at the extreme point of the distribution is 0, the density cannot be anything but zero. Let him tell me if I'm lying.
In MS Excel (the Russian version of Office 2010) there is the function gamma.rasp(), which when the parameters =GAMMA.RASP(0;1;1;0) returns an undefined value of the density at the point 0. Namely, it returns the error: #NUMBER!
Following Quantum logic, this point should be 0. Moreover, where the density is undefined, how can the integral on the left be defined?
However, when =GAMMA.RASP(0;1;1;1) (the cumulative density on the left is considered) I get a value of 0.
From this I draw a simple conclusion: either the assumption that for gamma at point 0 with these parameters Microsoft deceives us and its function cannot be trusted, or the density may not be zero and the integral on the left is still 0.
I asked Quantum a question, do you think Excel is wrong in estimating the density at zero too? Where is the answer?
In general, as a user, I am satisfied that I can trace the results of the algorithms in R by reading the authors' articles. I'm not satisfied with unsubstantiated claims and black-box algorithms in MQL, because I can't understand what some of their results are based on. As a user, you don't convince me.
It's just great. Already recognized mate layouts are not the truth. Do you believe in the code of the R optensor solution ....
There are rankings of statistical packages.
At the top are R, Pithon, SAS (paid) SPSS (paid). Of those you named, Matlab was in the ratings about five years ago, now it's gone.Wolfram (Mathematics) is not remembered at all. It, like Matlab, belongs to general math packages.
The paid version of R is called Revolution. After R was acquired by Microsoft, the division into paid and free remained.
Today, publishing articles on statistics without code in R or Python is considered indecent. And this is not by accident. Documentation on packages and functions in R necessarily has references to theoretical works that are implemented by these algorithms. The authors of these theoretical foundations for the algorithms are ALWAYS internationally recognized scientists. Because of this, R has become part of the global community of statisticians.
In my opinion, it is normal that the integral on the left at zero will be zero. And it is normal that the density will be infinite.
Such a heated discussion... Is there any practical sense in trading? What is the point of this research other than self-affirmation?