Iran an aw that - Page 3

supernothing · 23rd June 2009, 3:01pm

Just skimmed that aricle Les, but would you not need a bigger statistical sample than 116 numbers to run analysis of that type? I would think abberations in a sample that small would be quite likely. I'm more than willing to accept there is electoral fraud in Iran BTW, i just don't think it should be presumed from the off (as it was in the media).
On a related note: there was a lot of chatter about the "tell tale signs" of fraud - ie. the results being called so fast. Surely we should credit Amhedinejad with enough intelligence to not make this type of blatant error in hiding his theft - if theft it was. The guy is the leader of the most populous nation in the middle east, one of the most prominant voices in the developing world - he's not just some crackpot that goes "ah fuck it man just release the faked results now! I can't wait! The imperialists will never even notice!".

Also re: American hypocracy. I was reading (can't find the source) that there is still a couple of million dollahs in the American military budget that goes to fund Iranian terrorist groups destabalising the regime, which Obama hasn't changed.

supernothing · 24th June 2009, 1:51am

Aye I had a look at that article analysing the stats, surely that is nonsense? To take percentages like that from 116 results is bullshit - in a sample of 1000 you could examine that kind of thing, but such a small number must be very difficult.

AWESOMEUS MAXIMUS · 24th June 2009, 9:07am

More 'Pin the tale on the Ahmadinejad' from the bourgeois, IMO.
The more I get into this the more I think he may have actually won it fair and square, and it's upsetting America.

*LesMTS* · 24th June 2009, 9:16am

Quote:

Originally Posted by supernothing

Aye I had a look at that article analysing the stats, surely that is nonsense? To take percentages like that from 116 results is bullshit - in a sample of 1000 you could examine that kind of thing, but such a small number must be very difficult.

Well, I'm not a statistician, but 116 results is a fair number. It's not unusual for such sample sizes to be used. I've seen plenty of published epidemiological papers where n<50 for some situations.
You'd have to looks at exactly at what they did to see if it stands up to scrutiny (which isn't fully explained), but there's nothing about the fact that there's only 116 results that automatically makes it invalid. And a p value (<0.05) is given, which suggests that the results were corrected for sample size, and were still significant.

Semprini · 24th June 2009, 9:20am

Quote:

Originally Posted by AWESOMEUS MAXIMUS

The more I get into this the more I think he may have actually won it fair and square, and it's upsetting America.

"Ahmedinejad discounted the hanging gays chads illegally!"

endless psych · 24th June 2009, 9:46am

Well given that the population of results is 116...

It's valid aye.

TSR · 24th June 2009, 2:19pm

Quote:

Originally Posted by supernothing

he's not just some crackpot that goes "ah fuck it man just release the faked results now! I can't wait! The imperialists will never even notice!"

Unless he planned all of this so that he can promote anti-western ideals

Not saying that's what he's doing, that idea just popped into my head while reading your post

Quote:

Originally Posted by AWESOMEUS MAXIMUS

More 'Pin the tale on the Ahmadinejad' from the bourgeois, IMO.
The more I get into this the more I think he may have actually won it fair and square, and it's upsetting America.

He did appear to be more neutral towards America/the west recently. It's not him who deals with the Nuke Issue anyway

supernothing · 24th June 2009, 2:21pm

Quote:

Originally Posted by endless psych

Well given that the population of results is 116...

It's valid aye.

Can you explain please?

AWESOMEUS MAXIMUS · 24th June 2009, 2:38pm

Quote:

Originally Posted by TSR

He did appear to be more neutral towards America/the west recently. It's not him who deals with the Nuke Issue anyway

He is also not dealing oil on US currency, so I dare say that will be a major factor in their hate.

endless psych · 24th June 2009, 3:22pm

Quote:

Originally Posted by supernothing

Can you explain please?

There are only 116 numbers. The stats are testing the entire population of results not a sample, there are no larger numbers to be tested.

So in effect it's the population that the statistics are being run on and not a sample. If you hold that the basic contention (as follows) for doing this is correct:-

Quote:

We'll concentrate on vote counts -- the number of votes received by different candidates in different provinces -- and in particular the last and second-to-last digits of these numbers. For example, if a candidate received 14,579 votes in a province (Mr. Karroubi's actual vote count in Isfahan), we'll focus on digits 7 and 9.

This may seem strange, because these digits usually don't change who wins. In fact, last digits in a fair election don't tell us anything about the candidates, the make-up of the electorate or the context of the election. They are random noise in the sense that a fair vote count is as likely to end in 1 as it is to end in 2, 3, 4, or any other numeral. But that's exactly why they can serve as a litmus test for election fraud. For example, an election in which a majority of provincial vote counts ended in 5 would surely raise red flags.

Why would fraudulent numbers look any different? The reason is that humans are bad at making up numbers. Cognitive psychologists have found that study participants in lab experiments asked to write sequences of random digits will tend to select some digits more frequently than others.

Essentially you would expect a true election to show a more random distribution of numbers. Here there is a statistically significant result that suggests the numbers are not significantly different from one another.

supernothing · 25th June 2009, 6:44pm

I know the sample is the population and there are no larger ones available.
My contention is that with a field of only 116 numbers the randomness of the last digits is not a suitable test. On a much larger set of numbers you WOULD expect the last digits to take up their proportional share - but the smaller the set is the less suitable that test is. I would guess that 116 is just too small a set of numbers for this test to be relevant - if you randomly generated 116 numbers I would think it extremely unlikely that they would all appear equally.

**Shannow** · 25th June 2009, 6:48pm

This is mildly gigglesome.

http://idle.slashdot.org/story/09/06...athon?from=rss

Quote:

Iranian state television's Channel Two is playing a Lord of the Rings marathon in an attempt to keep people inside watching hobbits and not protesting in the streets. Normally people in Tehran are treated to one or two Hollywood movies a week, but with recent events the government hopes that sitting through a nine hour trilogy will take the fight out of most. Perhaps this was not the best choice in films if you want your people not to believe that "even the smallest person can change the course of the future."

*LesMTS* · 25th June 2009, 6:52pm

Quote:

Originally Posted by supernothing

I know the sample is the population and there are no larger ones available.
My contention is that with a field of only 116 numbers the randomness of the last digits is not a suitable test. On a much larger set of numbers you WOULD expect the last digits to take up their proportional share - but the smaller the set is the less suitable that test is. I would guess that 116 is just too small a set of numbers for this test to be relevant - if you randomly generated 116 numbers I would think it extremely unlikely that they would all appear equally.

What are you basing that on? It looks like you're just saying "it seems to me...". Of course larger samples are preferable but you can work out if the sample size is enough to give you valid results.
They've given a p value, which means they must have corrected for sample size, and the p value came out at less than 0.05. If the conclusion was invalid the p value wouldn't have been so low (although "significance" is selected more or less arbitrarily, 0.05 is pretty standard and pretty tight).
You'd have to see exactly what stats tests they used to be sure, like, and I'm no statistician.

endless psych · 25th June 2009, 6:59pm

With respect your objection is meaningless in terms of statistics. The size of a sample depends on the size of a population. The closer to the size of the population the sample the more powerful any statistical analysis is. Thus an analysis run on the entire population of results is the most powerful analysis you can do. Thus the one in two hundred change is valid and is statistically significant.

Because there are no other numbers to measure. There is also no need for more numbers. Cos that's not how statistics work. The rule isn't that an ever increasing absolute number makes a result more valid, or significant, but that a sample size close to the population, or one that increases relative to the size of the population becomes more valid.

Statistical significance may not imply substantive significance, you can take issue with the reasoning behind the process but you can't criticise the results by saying there aren't enough numbers without reinvening the field and theory of statistics. (but I you do want to that feel free to get rid of anything to do with logarithims! Cheers

)

supernothing · 25th June 2009, 7:29pm

Might seem like i'm being pigheieded but i'm not meaning to be, BUT

Quote:

Originally Posted by endless psych

With respect your objection is meaningless in terms of statistics. The size of a sample depends on the size of a population. The closer to the size of the population the sample the more powerful any statistical analysis is. Thus an analysis run on the entire population of results is the most powerful analysis you can do. Thus the one in two hundred change is valid and is statistically significant.

The fact that the sample size=the whole population (aside from the fact that for some reason the statisticians never included the "spoiled ballots" as another candidate which would have increased the sample size 20%) doesn't really matter with this type of analysis does it? Its to do with the amount of actual results (in this case only 116) - if they had smaller constituencies for exaple we would have more results and the test of whether the last digit is "random enough" would gain weight.

Quote:

The rule isn't that an ever increasing absolute number makes a result more valid, or significant, but that a sample size close to the population, or one that increases relative to the size of the population becomes more valid.

I would think that WAS the rule in this type of analysis - that programme that runs the "monty hall problem" comes to mind, the more times you simulate the choice the more you believe the outcome.