Urgent! Real quick question! Can we make a conclusion/inference from this graph?

riseofphoenix

This is what I did on StatsCrunch using the data they gave me:




Would the answer be B and C?
Because A can't be right...our data goes up to 2000. And while there still is a trend going on (newer homes --> more square ft), we can't really give a "predicted value of 1964 and 2250". So that would eliminate D and E (the fourth option and the sixth option).
Am I right? I only get one submission and I have to choose AT LEAST ONE of these.

Number Nine

Explain clearly why you rejected the options that you did. I can see an obvious problem with the dataset that you're ignoring.

riseofphoenix

Option 1 is false because the the scatter plot isn't necessarily linear - it's pretty randomized
Option 2 could be true because our parameter is 1920-2000
Option 3 is true.
Options 4, 5, and 6 are false.

Number Nine

Explain why.

riseofphoenix

Well you can't predict exactly how many square ft. a home built 15 years from the year 2000 will have. Especially considering the limited data that was given.
Option 5 just doesn't make sense to me. So it has to be Options 2 and 3.

Number Nine

It "doesn't make sense"? Do you understand the assumptions involved in least squares regression? Does the dataset look like it has equal variance across observations?

riseofphoenix

Oh well not /completely. But to answer your second question, no it doesn't.

So Option 5 would also be the correct answer (along with Options 2 and 3)?

haruspex

That's a fair point, but the question does not specify the kind of analysis to be used. Is it not possible to make allowance for variable variance (heteroscedasticity)?

chiro

I'd really want to see an R^2 value for this data because it looks absolutely unusable to say the least given the scatterplot.

If the variables are high un-correlated, then any attempt to create dependencies between the variables is going to be useless.