Regression Analysis Tools

[Metamorphic]

I'm consistently underwhelmed by regression analysis for most appraisal purposes. Where you have a sufficiently large and consistent population that regression works well, you really don't need it to make a reliable value conclusion. When you're dealing with high value and exotic amenities that are the most difficult to value you never have enough data to make regression work.

 

[erockinaz]

I would learn how to develop regression analysis and then use Excel. I also use Gnumeric, a lot!!. Its free, has powerful statistical tools, including a powerful regression analysis tool. Its important to learn regression analysis to fully analyze the data, especially if you are going to include it in an appraisal.

 

[George Hatch]

I'm consistently underwhelmed by regression analysis for most appraisal purposes. Where you have a sufficiently large and consistent population that regression works well, you really don't need it to make a reliable value conclusion. When you're dealing with high value and exotic amenities that are the most difficult to value you never have enough data to make regression work.

^^^ what he said

 

[Terrel L. Shields]

I'm no rocket surgeon but $600 a year is a hefty price for a simple regression program. I think Eugene Pasymowski only gets half or so that for RealStat and its good for years.

 

[Elliott]

Free is a pretty good price. Even if you pay for your analysis, I'm willing to bet most appraisers can't understand it or explain it, or tie it to their market.

http://www.dummies.com/how-to/content/how-to-calculate-a-regression-line.html

https://www.easycalculation.com/statistics/regression.php

http://www.mathportal.org/calculato...tor/correlation-and-regression-calculator.php

 

[George Hatch]

I am of the opinion that if you can't explain what the tool is doing or replicate what it's doing on your own then you shouldn't be using it.

The equations for such an analysis are dirt simple:

maximum (of your range)
- minimum (of your range)
= difference (in your range)

You're going to start with one adjustment factor (doesn't matter where) and continue to try different factors until you find the factor that is most effective at reducing that difference.

If you were going to use this process for adjusting for several attributes you'd want to prioritize which attributes are of the most effect and start with those.

The Excel syntax for the above equations are:

 

[Kim Whiting]

and you need a regression analysis program to do that simple process?

 

[Terrel L. Shields]

Sensitivity analysis is basically multiple linear regression for one isolated item

 

[George Hatch]

It's a variation of the process many appraisers already use when developing adjustment factors with a dataset. You try out different factors to see which ones work best to produce that convergence among your adjusted value indicators. The big difference is you're usually doing it manually.

All a spreadsheet adds to the process is the ability to do this with larger datasets and the ability to quickly (and accurately) do multiple iterations. The point remains that you're extracting the adjustment factors you're using from the sales data that includes those comps you're actually trying to adjust.

If the adjustment factor that works best for the BIG DATA dataset is $50/sf but your comps are reacting better to $100/sf then it doesn't matter what the larger group is doing because you're not adjusting to them, you're adjusting to your direct comparables. That makes the $100/sf you extracted from your dataset the optimum factor for your comps.

 

[Michael Willis]

I'm consistently underwhelmed by regression analysis for most appraisal purposes. Where you have a sufficiently large and consistent population that regression works well, you really don't need it to make a reliable value conclusion. When you're dealing with high value and exotic amenities that are the most difficult to value you never have enough data to make regression work.

I completely agree. In the area I service, I find this is the case more times than not. I still do regression analysis on almost every property but report why the results were unreliable or not. With that said, I am pretty comfortable with Excel and Minitab since I used it in all my finance classes, I don't see myself shelling out extra dough for regression programs/tools. Most MLS systems allow users to import data into Excel. Play around with it, it's pretty straightforward to use (interpreting the data and watching for outliers is where it gets tricky).