Am a beginner in data analysis and i feel that something about these graphs are wrong. Am not sure that the outliers are the issue or am doing this the wrong way. Thanks for the help Scatterplot

  • 1
    $\begingroup$ Can you provide some more information - 1. What is your outcome variable? (what are its units?) 2. Which model did you use? 3. Have you performed any model diagnostics? $\endgroup$ – NB12 Jun 27 '19 at 6:28
  • $\begingroup$ This is just plotting area of property to predict the price of that property. The 4 different graphs represents 4 different area plotted against the price of property, which are LivingArea, BasementArea, 1stFloorArea, and GarageArea respectively. I have not build any model around it yet. $\endgroup$ – Aaron Jun 27 '19 at 6:52
  • 2
    $\begingroup$ Are you concerned about the changing spread of data about the line or something else? Note that when expected house price is 0 you'd expect the variance to also be 0 - so changing spread with changing expected price is hardly a surprise. I'd probably lean toward some form of gamma model as a first thought. $\endgroup$ – Glen_b Jun 27 '19 at 7:52
  • $\begingroup$ unrelated but if you are trying to predict housing price data for any commercial reason you may want to read money.stackexchange.com/questions/110537/… from yesterday $\endgroup$ – MD-Tech Jun 27 '19 at 14:17

It doesn't seem wrong. The linear fit passes through the data reasonably well. If you're concerned about the spread around the fitted line in the region with less data, it actually reflects the uncertainty in the model around different domains; which is typically calculated via Bayesian Linear Regression.

| cite | improve this answer | |
  • 4
    $\begingroup$ I wonder how relevant, valid and/or correct the point about Bayesian linear regression is. $\endgroup$ – Richard Hardy Jun 27 '19 at 13:07
  • 1
    $\begingroup$ @RichardHardy have you done some research to you find out if there's an issue? $\endgroup$ – David Jun 27 '19 at 13:48
  • 3
    $\begingroup$ @David, ideally, I would have. However, given the limited resources, I chose not to. There is also the question of who should bring the evidence: one that is claiming a fact without supporting evidence or one that does not quite believe it? I would tilt towards the former. Based on my current state of knowledge, I have some doubts that I have politely expressed in the form of "I wonder", just to give the author a heads up. Perhaps the statement was not thought through and the author would change it. Or if it is correct, then it would be reassuring to at least get a brief confirmation. $\endgroup$ – Richard Hardy Jun 27 '19 at 14:10

Not the answer you're looking for? Browse other questions tagged or ask your own question.