Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

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

  1. Nov 14, 2012 #1

    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.
  2. jcsd
  3. Nov 14, 2012 #2
    Explain clearly why you rejected the options that you did. I can see an obvious problem with the dataset that you're ignoring.
  4. Nov 14, 2012 #3
    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.
  5. Nov 14, 2012 #4
    Explain why.
  6. Nov 14, 2012 #5
    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.
  7. Nov 14, 2012 #6
    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?
  8. Nov 14, 2012 #7
    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)?
  9. Nov 15, 2012 #8


    User Avatar
    Science Advisor
    Homework Helper
    Gold Member

    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)?
  10. Nov 15, 2012 #9


    User Avatar
    Science Advisor

    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.
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook