Statistics question - bleh!!

[a_egan9]

Hey I need sum help wif a Research Methods assignment. The question is if two variables are totally independent, then the correlation between them is:

 

1) -.01

2) -1

3) +1

4) zero

5) none of the above

 

Any help would be great Im very stressed!!!

[5614]

Well you seem to be asking what correlation is. And exactly what the question is saying.

 

The question says if two variables, that means there are 2 things and they change, are independant, that means they do not effect each other. For example if you throw a dice it's always a 1/6 chance, no matter what you got the time before, however if you are picking things out a bag and do not replace then the probability changes. So say there's 5 chocolate bars and 5 sweets and I pick a sweet the chances of getting choc is now 5/9 (because there's 1 sweet missing) whereas before it was 5/10.

 

In this question there are two things which change which do not effect each other, like rolling 2 dice.

 

Then what is correlation. Correlation is the relationship between 2 things, variables in this case.

 

A correlation of +1 means that when one variable is something the other is the same. So for dice rolling 6 on both would have a high correlation.

 

A correlation of -1 means the when one variable is something other is the opposite. So like a 6 and 1.

 

No correlation or 0 correlation means that it's random, sometimes one is high the other is low, sometimes theyre the same, it's all jumbled up and random.

 

So which one is it? The two variables do not effect each other, but that doesn't mean much! They might be the same every time, they might be opposite, they might be random, we do not know.

 

So I'd say the answer has to be #5 "none of the above" because the question does not give us enough information to guess the true correlation.

[a_egan9]

thanx i thought it would be either zero or none of the above but i guess i didn't really understand the question (or the theory behind it)

[a_egan9]

is it possible to calculate effect size if u hav pearson's r value?

[Klaynos]

Sorry to differ 5614 but if they are "totally independent" then there cannot be corrolation, no corrolation = 0

[5614]

Sorry to differ 5614 but if they are "totally independent" then there cannot be corrolation, no corrolation = 0

Is that what it means??? It doesn't mean that, well, like I thought, independant as in one does not depend on the other, but means that they have 0 correlation. You sure? (Not disbelieving you, but just I've never heard of it and I know what independant means!)

 

I know it's random and it's independant, but because it's random there is a possibility that they could be correlated, or, is my definition of "totally independant" wrong?

[Klaynos]

Well if two things are completely unreleated, how can their be corolation between them? There can't be therefore corrolation must be 0, if there is something relating the two variables then there has to be some form of corrolation even if it is very low...

[5614]

There might be though! Throwing 2 dice is independant of each other. The chances of getting 6 on both is 1/36 and the chance of getting that twice would be tiny, but that doesn't mean it can't happen.

 

Now we do not know the situation here, only that there are 2 seperate random events. Surely there's a possibility no matter how small that there could be a correlation, if it's truely random it is possible!

[Klaynos]

yes but the probably of getting:

 

6 - 6

 

Is exactly the same as getting

 

6 - 1

 

or ANY other predefined pattern...

 

2 seperatarate independent events have 0 corrolation.

[5614]

What's your point?!?

 

Every 36 times you roll you will get 6-6... every 1296 times you roll you will get 6-6 & 6-6.... so say this is being counted over 5 rolls then once every *massive number* you roll you will get 5 lots of double 6.

 

Because of this we cannot be certain, sure they are random, they may not be correlatded, but, like I said above, they may be, we don't know.

[Glider]

Where two variables are completely independent, there will be no correlation. Rolling a pair of dice and recording the values will, after a sufficient number of trials, yield no corelation (r = 0 or a value close to zero).

[ku]

As I recall reading from Flying Spaghetti Monsterism, average global temperature and the number of pirates in the world is negatively correlated. These two variables are independent.

[Klaynos]

is it possible to calculate effect size if u hav pearson's r value? [/b']

 

Hi sorry but it's not a value I've come accross enough to know

 

What's your point?!?

 

Every 36 times you roll you will get 6-6... every 1296 times you roll you will get 6-6 & 6-6.... so say this is being counted over 5 rolls then once every *massive number* you roll you will get 5 lots of double 6.

 

Because of this we cannot be certain' date=' sure they are random, they may not be correlatded, but, like I said above, they may be, we don't know.[/quote']

 

 

If they are random and you do it enough times there will be no corrolation, I am right simple. Sorry if that sounds big headed, but this is just something I know.

[AL]

The covariance of two independent random variables X and Y is zero.

Cov(X, Y) = 0

 

The correlation coefficient of two random variables X and Y is:

p = Cov(X, Y) / (Var(X)*Var(Y))^0.5

 

Thus if two random variables X and Y are independent, their covariance is zero, and so is their correlation coefficient p.

[ku]

If variables are independent, they are uncorrelated. If variables are dependent they may be uncorrelated.

[a_egan9]

thanx guys if anyone has any clues on the other question that has been puzzling me any help would be greatly appreciated!

 

* someone who runs a correlational analysis says that an effect size of 64% has been found. What value of Pearson's r did they obtain?

 

1) +0.8

2) -0.8

3) 0.8, we cannot tell whether the value is positive or negative

4) +0.64

5) none of the above

[5614]

If they are random and you do it enough times[/b'] there will be no corrolation

That with emphasis on the bold is totaly true and is the angle you are coming from!

 

I was saying if you didn't do it 'enough' times and you were in a, statisticly speaking, highly improbably position of having a few 6-6s in row (on two dice).

 

But your "enough times" is very true. After all each repetition will, on average, make the average of your data more acurate. So yes, you are right... (if you didn't do it enough times and the very improbable happened, but hey, lets not go into that!)

 

As for the correlation/pearson's rank question I don't know how to do that, you may get an answer if you ask in the Maths forum section.

[Klaynos]

That with emphasis on the bold is totaly true and is the angle you are coming from!

 

I was saying if you didn't do it 'enough' times and you were in a' date=' statisticly speaking, highly improbably position of having a few 6-6s in row (on two dice).

 

But your "enough times" is very true. After all each repetition will, on average, make the average of your data more acurate. So yes, you are right... [i'](if you didn't do it enough times and the very improbable happened, but hey, lets not go into that!)[/i]

 

As for the correlation/pearson's rank question I don't know how to do that, you may get an answer if you ask in the Maths forum section.

 

Well if you didn't do it enough times and because of that found a corrolation, you're results are flawed, and any conclusions you make are meaningless

[5614]

Well if you didn't do it enough times and because of that found a corrolation, you're results are flawed, and any conclusions you make are meaningless

That is of course correct. However we don't know the circumstances, maybe it is not as simple as rolling some dice, maybe the people doing the experiment don't have a clue whethere there will be a correlation or not, if this was the scenario and they didn't have enough repeats and had highly improbable results then they could find a correlation in a place where there shouldn't be. This is where I was coming from.

[BigMoosie]

5614: There is absolutely no correlation between to independent variables. You are thinking in a more practical sense where in the real world basically everything is correlated to the infitessimal degree, but for all mathematical purposes the answer is zero.

[Klaynos]

We're told they are totaly independent in the question, if this fact is ignored then you might as well ignore the rest of the question and do a bit of fourier analysis...