Linear Regression: Counts or Percents

[zackb11]

I am running a linear regression to see if there is a relationship between the slope of a hillside and the number of rocks of a certain size on that hillside. My first variable is slope. Should the second variable be the NUMBER of a particular rock, or the PERCENTAGE rocks of that particular class in comparison to my entire sample. ie: I have a sample of about 80 rocks, if 20 of them are class A, should I use the value of 20, or should I use the value of 25%. This analysis is being done at several locations along the hillside. Thanks

--Zack

 

[tkhunny]

That choice would depend on your experimental design. Did you design your experiment and WRITE IT DOWN before you started collecting rocks? That would have been the thing to do. If not, you must decide, but likely it is of no consequence, you will just get a different relationship, but not come to a different conclusion.

Using 20: 20(k) = Slope = 8% Grade ==> k = 8%/20 = 0.004
Using 25% 25%(k) = Slope = 8% Grade ==> k = 8%/25% = 0.32

You just get a different constant (k), it doesn't make the problem any different.

Next time, write your plan better and more in advance. :wink:

 

[zackb11]

should my slopes be in degrees or percentages? or does that not matter too. I'd prefer to use degrees and counts, for slope and rock samples, respectively, because that's what I have done already. I'm trying to meet a deadline with this project.

 

[tkhunny]

1) Write things down in advance. Then you will not be confused.

2) Do it in a way that makes sense to you and can be communicated to others. Then no one will be confused.

3) Please make sure all the units match up. If you start talking about percents and wind up talking about degrees, you will lose some of your audience.