1
$\begingroup$

I have the following problem. I am trying to use the poisson regression on my data, which look like this:

Year   Resistance.proportion
1990            0.367
1991            0.678
1992            0.786

I am using the following code in R to do the modeling:

> model <- glm(Resistance.proportion ~ Year, data=sulfodf_for_poisson_r, family = poisson(link=log))

After executing the command, I get the following message in R:

Warning messages:
1: In dpois(y, mu, log = TRUE) : non-integer x = 0.362069
2: In dpois(y, mu, log = TRUE) : non-integer x = 0.375000
3: In dpois(y, mu, log = TRUE) : non-integer x = 0.723684
4: In dpois(y, mu, log = TRUE) : non-integer x = 0.458333
5: In dpois(y, mu, log = TRUE) : non-integer x = 0.595238
6: In dpois(y, mu, log = TRUE) : non-integer x = 0.666667
7: In dpois(y, mu, log = TRUE) : non-integer x = 0.875000
8: In dpois(y, mu, log = TRUE) : non-integer x = 0.583333
9: In dpois(y, mu, log = TRUE) : non-integer x = 0.321429
10: In dpois(y, mu, log = TRUE) : non-integer x = 0.933333

Is the problem here really that the numbers are floats and have to be integers or is there something more subtle? Furthermore, if I summarize the model, I get an AIC that is infinite, which is also a bad thing as I understood.

I have looked through several questions as Fitting a Poisson GLM in R - issues with rates vs. counts, but they do not provide an helpful answer to me.

So my questions are:

1.)Why am I getting these errors and how can I resolve them?

2.) Once the model is correct, what is a good way to graphically represent it?

$\endgroup$
4
  • 3
    $\begingroup$ Your title says 'count data' but the example data seems to contain proportions rather than counts $\endgroup$ Commented Apr 4, 2016 at 11:36
  • 2
    $\begingroup$ You could try beta regression instead of Poisson. $\endgroup$
    – Peter Flom
    Commented Apr 4, 2016 at 11:42
  • $\begingroup$ Oh man...that was a pretty stupid error. yes, transformed the count data into proportions. $\endgroup$ Commented Apr 4, 2016 at 11:44
  • $\begingroup$ @PeterFlom thanks, i will read up on beta-regression! $\endgroup$ Commented Apr 4, 2016 at 11:49

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.