# correlation question

• Jun 26th 2013, 02:49 AM
pumbaa213
correlation question
I'm not sure if I should have posted this under statistics or advanced statistics so if this question is placed incorrectly, I am sorry.

The question shows 2 scatter diagrams
I can't upload the pictures as Im using my mobile but i'll just describe how they kinda look

scatter diagram 1: the points form a very straight line
scatter diagram 3 : the points form a somehow straight line but not as straight as scatter diagram 1
Both scatter diagrams do not have statistics given

question
the two scatter diagrams are for 2 samples of bivariate data. The scales on each axis are identical

(i) for the sample shown in scatter diagram 2 , state an approximate value for the linear (product moment) correlation coefficient r.

is there a way to answer this question? there is no statistics given on the scatter diagram so I think I am supposed to just estimate r based on how the scatter diagram looks like?

(ii) the value of r for the sample for scatter diagram 1 is + 1. Explain why this need not imply that a linear relationship holds for the whole population.

I hope you'll be able to understand what I wrote If there are any uncertainties I Will try my best to clarify.

Thank you very much!! :)
• Jun 26th 2013, 02:51 AM
pumbaa213
Re: correlation question
Sorry typo error

Scatter diagram 1:
Scatter diagram 2 (instead of 3)
• Jun 26th 2013, 10:34 AM
DerWundermann
Re: correlation question
Not sure if this is going to be useful, but:

(1) Difficult to answer that without the figure. But I guess that rho would be roughly 1
(2) It seems simple: The rho you are referring to is a sample correlation, so the results are valid for the sample at hand and that rho is only an estimate for the population correlation.
That estimate, however, is associated with a certain degree of uncertainty, which reflects the sample size available. The smaller the sample size, the larger our uncertainty.
To sum up, even if the sample correlation rho is 1, there is uncertainty in relation to the population correlation.