1
$\begingroup$

I want to do a linear regression for my data but I’m not sure what I need to do with my raw data before running a linear regression. I have counts for species in several forests of differing sizes. I want to compare these counts to percentages of a certain land type within a fixed region around the forest. My worry is that large forests will have large species counts and won’t reflect the land type which is the whole purpose. How can I standardize the count data (is it even standardizing I should be doing?)

$\endgroup$
  • $\begingroup$ Welcome to CV. Your question isn't completely clear. What is your dependent variable? What are your independent variables? $\endgroup$ – Peter Flom Dec 7 '18 at 12:39
  • $\begingroup$ @Peter Flom Sorry about that. My independent variable is percentage of developed land within a fixed distance around a forest. The dependent variable is the number of trees of a certain species (count). I have multiple forests of different sizes, some small with low counts and some large. I’m assuming I can’t just plot percentage and counts on a graph as they are. $\endgroup$ – NeedtoPass Dec 7 '18 at 15:30
1
$\begingroup$

You should include an offset term. With count data I'm assuming you're doing some kind of poisson or negative binomial regression.

So, for example, using the formula for a negative binomial regression with a log link:

$$\ln(\mu_i)=\beta^tx_i$$

$$\mu=e^{\beta_1x_1+\beta_2x_2+\epsilon}$$

adding an offset for size of forests makes it

$$\ln(\frac{\mu_i}{size_i})=\beta^tx_i$$

$$\mu=e^{\beta_1x_1+\beta_2x_2+ \epsilon}+size$$

Size gets added in 'raw' it doesn't have a beta. The interpretation is now the expected number of species per unit of size of forest is whatever.

$\endgroup$
  • $\begingroup$ my notation's very bad, but i hope it still conveys the message $\endgroup$ – Huy Pham Dec 7 '18 at 15:43
  • $\begingroup$ I understand what you mean and that looks like the right way to go about it. The only problem is I don't have the size of each forest delineated in my dataset. I will try replacing size with number of sampling sites within in each forest and see if that works. Thank you! $\endgroup$ – NeedtoPass Dec 7 '18 at 16:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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