PDA

View Full Version : Random thoughts : Dilbert 10/25/2001

Actuary321
10-25-2001, 01:38 PM
http://www.dilbert.com/

Today's Dilbert was really funny in light of our recent audit.

The auditor had generated a set of random numbers to use to pick policies that he would look at. He then came to me asking if I knew of a way in Excel to generate the same set of random numbers again so that he would be able to show his supervisor that he had indeed generated a set of random numbers.

Dr T Non-Fan
10-25-2001, 01:47 PM
I was thinking about collecting these, putting them up where the accounting/financial folks can see them.

"Go ahead, spit on his data." (Yesterday's.) That particularly hits home. F'in idiots.

mackinac44
10-25-2001, 02:46 PM
What's even funnier is that the auditor is right.

I just generated the following two sets of 10 random numbers using the built-in feature in Excel and got identical results.

If you use the random number generator for a Uniform distribution with a seed value of 3, you'll probably get the same results.

Run #1 Run #2
0.001464888 0.001464888
0.219611194 0.219611194
0.283639027 0.283639027
0.27744377 0.27744377
0.21457564 0.21457564
0.719534898 0.719534898
0.540238655 0.540238655
0.717276528 0.717276528
0.830622272 0.830622272
0.371349223 0.371349223

<font size=-1>[ This Message was edited by: mackinac44 on 2001-10-25 14:47 ]</font>

Steve White
10-25-2001, 03:31 PM
Perhaps that depends on your version of Excel. In Excel 97, the Rand() function does not seem to take an argument or seed.

You could use the rnd function in a macro, with a seed, to generate the same random numbers many times.

Neither seems relevant to the auditor's request. He had already generated his random numbers (not necessarily even in Excel). He asked for way to reproduce those same numbers in Excel, using Excel's random number generator.

mackinac44
10-25-2001, 03:47 PM
I didn't use a function to generate the random numbers. Excel has a built-in random number generator. It's part of a Data Analysis add-in that has to be manually activated by the user.

To activate the Data Analysis option, first select TOOLS, then ADD-INS, and check the Analysis Toolpak. The Data Analysis option should then appear as part of the TOOLS menu. One of its options is random number generation.

Dr T Non-Fan
10-25-2001, 04:21 PM
So, there's an add-in that has the random numbers, and in some kind of order?

In that case, you'll need to randomize the seed, eh? Perhaps this is what the instructions (see if you can find them) instruct?

Steve White
10-25-2001, 04:28 PM
Neat.

I didn't have that add-in enabled before. With it, I get your same 10 numbers.

That still doesn't help match a set of numbers (whose source is unknown) by a procedure that can be demonstated as generating random numbers.

Dr T Non-Fan
10-25-2001, 04:57 PM
Seems as if it were created so that random processes CAN use some official random number table. Like those found in the back of statistics books.

As if to say to an auditor in a fancy memorandum, "We used the Excel random number processor. Our seed was 4,951, which was generated by round(Rand()*10000,0)."
So that the auditor could check the results. Using this list of officially approved random numbers, and starting your process at some random (but documented) seed point allows the the process to be audited, as well as allows it to be certified as random, if that's what's needed.

As if some group of auditors got together and asked Microsoft to come up with this.

Snoopy
10-25-2001, 05:24 PM
I believe the seed is just the time of day (in seconds or something like that).

Actuary321
10-26-2001, 02:20 PM
Sorry, I made it sound as if the auditor had generated the list on his own some other way. In fact he had used Excel in the way described above, and was asking both how could he reproduce the list of random numbers (which he figured out the way that was described above) and how those number could then be random, if the exact same set could be generated again.

WQN
10-26-2001, 02:34 PM
Any pseudorandom number generator, as found in Excel and most anywhere else, will generate the same list with the same seed. This is part of the algorithm. Its illogical to think that a software application can produce anything truly random.

Mr. Grim
10-26-2001, 02:37 PM
Are you telling those there slot machines I play are not truly random? Where is Trump I am going to kick his ass?

Steve Grondin
11-03-2001, 08:54 PM
One of the most intellectually intriuging scams against casinos that I have heard of takes advantage of the fact electronic slots must use a pseudorandom generator. The perpetrator knew the "randomizer" formula. He would play a small amount until he knew a winner was coming up next and would hike the bet. I forget how successful he was before he was found out. Now the casinos pay attention to betting patterns like this just in case.

Rosebud
11-05-2001, 12:25 PM
Steve,

While the slots do need to use a pseudorandomizer, they cycle at a rate of thousandths of milliseconds, so you cannot control the seed number that way.

Don Quijote
11-05-2001, 05:16 PM
assumption 1: the auditor doesn´t really care if the numbers can be reproduced, he just wants some documentation.

method 1: write a macro in Excel to create a list of numbers, which you type in from his random list. name the macro rand1 or something similar. Document that "using the function rand1, and a seed value of X, the following list of random numbers was generated, which were then mapped into policy numbers using the following procedure."

method 2: tell the auditor that he´s crazy and that nobody up the chain will ask if the policies were selected randomly and even if they do, an adequate answer is "yes".

Steve Grondin
11-06-2001, 02:55 PM
Rosebud,

The source for my info was a TV show, so of course they didn't go into the mathematical detail. To elaborate a little more, the guy was in on the programming of the machines, so he knew what the formula was. The formula was such that if a specific sequence came up, he knew what was coming next. It seems reasonable to assume the casinos have changed the formulas so there is not a fixed progression, but one that depends on the time the button is pressed, making the process appear more random (and harder to cheat).

Darth Tater
12-03-2001, 02:25 PM
As a professor of mine in grad school once said, nothing is truely random. Just run these thoughts by the Law of Large numbers and see what you get.

Anonymous
12-04-2001, 07:57 PM
Every random number generator uses a seed. What that seed is and how that seed is selected is another matter. Sometimes one can arbitrarily define the seed.

If anyone has an HP15C or HP12C you'll find that you can reset the calculator to its original "out-of-the-box" settings to always get the same first random number after each and every reset.

Anonymous
12-04-2001, 07:59 PM
Thus, this is the essence of

"God does not shoot craps."
vs.
"God gave us free will."
vs.
"Everything happens for a reason."

I've gotta get outta these pants!

Drzy
12-05-2001, 06:02 PM
Does anyone have a picture of the Dilbert cartoon still? I can't find it on the internet anymore. I'm doing a presentation on random numbers, and that would fit in very nicely with my PowerPoint slides. :smile:

Dr T Non-Fan
12-05-2001, 06:33 PM
Go to the dilbert website in the original post.

In the middle of the web page there's an area that states:
" Posters: Get any Dilbert strip published in the last 90 days printed on a poster! Selected older strips are also available."

It will open a new page. Then enter the date 10/25 into the date choice.

I'd tell you the address, but I'd have to kill you first. Haha. Actually, the address bar wasn't showing.

Dr T Non-Fan
12-07-2001, 03:07 PM
Someone's forgotten their manners.

Drzy
02-12-2002, 11:45 PM
Oh, crap!! Sorry, this is way way late, but thank you very much! I'm starting off the presentation with a fake random number generator that'll give me a bunch of nines, and I'll fake confusion, say it's worked all the other times, and then show the Dilbert cartoon. It should go over great. :smile:

Again, a huge serious thank you.

Dr T Non-Fan
02-13-2002, 12:17 AM
See if I ever help ANYBODY EVER AGAIN!