The new Primary Thread

[Dinner]

The Black Eyed Peas have endorsed Obama and made a nice video/support piece for him.

[Cyclotron]

Damn

[Kuciwalker]

Originally posted by MarkG
i understand that, but isnt the US more diversified than the average country and thus needing a bigger sample?

That doesn't have any influence on the math.

edit: btw, I'm not disputing anything snoopy says - I don't know the SE of various sample sizes off the top of my head

[OzzyKP]

Originally posted by snoopy369
I like the refusal to apologize for the war vote I prefer a politician who will stick with what s/he believes in, and change his/her mind when the situation changes but not claim "oh, I was wrong before". No, you did what you felt was right then, and now due to new information/circumstances this is no longer the case.

Yea, but many of us want a leader who will make the right decision on day one.

[snoopy369]

Originally posted by Kuciwalker


That doesn't have any influence on the math.

edit: btw, I'm not disputing anything snoopy says - I don't know the SE of various sample sizes off the top of my head

Actually, it absolutely does influence the SE, when calculated right. Or, more accurately, it likely influences the SE.

The error of a sample comes from the effective sample size, not the actual size. The effective sample size is a measure of how close the sample distribution comes to the population. A group of 1000 people who are mostly 60+, for example, might have an effective sample size of 300 - meaning that it has approximately the same predictive power as a sample of 300 people that exactly reflects the population.

Effective sample size is a very approximate science, unfortunately - you have to choose which elements to base your guess as to how close the sample is to the population, as you can't pick EVERY variable. Typically you'd pick four or five categories. Age, gender, income level, party affiliation, and race, for example, might be one set of criteria. You try to pick criteria that are both easily measured, and accurately define the potential differences; you could measure height, for example, but more than likely height is not highly correlated to who you will vote for Education, MSA/County/other geographical indicator, Employment, Housing status, etc. are all variables you could use; but use too many, and it ends up overly complicated.

The more the population varies in the criteria selected, the lower the effective sample size will tend to be. In theory you could pick a perfect sample of 1000 people and have an effective sample size of 1000, regardless of the population; but in practice, the more variation in those variables in the base population, the harder to pick a good sample (and less likely it is to pick one by random selection), particularly when the selecting process has an effect itself (people more likely to be at home, people more likely to volunteer to answer, people more likely to have a telephone line, etc.) A perfectly homogenous population (10 million white male Democrats of age 35 all making $70k a year) would give you a perfect sample size of 1000; any deviation from that reduces your likely effective sample size.

[snoopy369]

Originally posted by Cyclotron


I only like stubbornness when it's stubbornness in support of the right decisions.

One of my biggest issues with the Bush administration was the complete inability of anybody to admit to mistakes or take responsibility for anything. The way I see it, if you think the war wasn't a mistake, you're a fool, and if you think it was a mistake, then you should apologize if you originally voted for it. There is no middle ground, despite Hillary's refusal to admit it. I respect a politician changing their mind and admitting past mistakes much more than simple pigheadedness.

I'm not talking stubbornness; I much prefer a politician who changes his/her mind when there is new information that affects the decision. However, I don't like it when politicians say, "At the time, I should have voted the other way"; based on what knowledge they had in 2002, they should have voted the way they did. Perhaps the Bush administration knew better at the time; but the Senate didn't, certainly. Senators who come out now and say, "I should have been against the war in 2002" I have nothing but disdain for. I have respect for those who say, "Based on what I knew in 2002, I supported the war; I no longer support it based on what has happened and what I know now".

[Arrian]

Perhaps the Bush administration knew better at the time; but the Senate didn't, certainly.

Why so certain the Senate didn't know better? FFS, I knew better, and those folks are supposed to be better informed. Key word supposed.

There were those in the Senate who saw things clearly back then and made the correct decision in spite of the potential political toxicity of a nay vote. Unfortunately, none of them are presidential candidates.

-Arrian

[Dinner]

So am I the only one who liked that Obama piece?

[Kuciwalker]

Originally posted by snoopy369
*snip*

In theory you could pick a perfect sample of 1000 people and have an effective sample size of 1000, regardless of the population;

*snip*

I probably should have mentioned that this was my assumption (since textbooks always use it). I know it's not a very good one

[Kidlicious]

Originally posted by Oerdin
So am I the only one who liked that Obama piece?

It's kind of weird for some reason.

[snoopy369]

Originally posted by Kuciwalker


I probably should have mentioned that this was my assumption (since textbooks always use it). I know it's not a very good one

Indeed it's not, and it rather undermines your argument

[Flubber]

Originally posted by DinoDoc
If that analysis is correct, why doesn't one of them quit?

Perhaps because each of them would believe, deep down, that THEY were their party's best chance for victory

[Ben Kenobi]

The more the population varies in the criteria selected, the lower the effective sample size will tend to be. In theory you could pick a perfect sample of 1000 people and have an effective sample size of 1000, regardless of the population; but in practice, the more variation in those variables in the base population, the harder to pick a good sample (and less likely it is to pick one by random selection), particularly when the selecting process has an effect itself (people more likely to be at home, people more likely to volunteer to answer, people more likely to have a telephone line, etc.) A perfectly homogenous population (10 million white male Democrats of age 35 all making $70k a year) would give you a perfect sample size of 1000; any deviation from that reduces your likely effective sample size.

Yes, that's very true. If your sample has way more young people, then it's not going to predict well what the overall vote should be in the election.

[snoopy369]

Most surveys are heavily biased towards older people, on balance, because older people:
a) are at home (retired)
b) like to take surveys
c) don't screen their calls

It's certainly possible for surveys to be biased towards young people, but you pretty much have to try. In-person surveys might be more so, and paid surveys as well - but in general, phone surveys like this tend towards the 65+ demographic.

[Flubber]

Originally posted by snoopy369
Most surveys are heavily biased towards older people, on balance, because older people:
a) are at home (retired)
b) like to take surveys
c) don't screen their calls

It's certainly possible for surveys to be biased towards young people, but you pretty much have to try. In-person surveys might be more so, and paid surveys as well - but in general, phone surveys like this tend towards the 65+ demographic.

Agreed-- Isn't there a big bias in a LOT of phone surveys since they typically use 'land-line' phone numbers and don't tab into the people that use cells only?