0
$\begingroup$

I'm trying to see the isolated and combined effects of precipitation and three different human disturbances indexes are having an effect on the growth of a cactus. I have a gradient of eight different values of precipitation and eight different values of each index (livestock grazing, wood extraction, and people pressure). The problem is I only have one value of growth per each precipitation and disturbance index, so when I run the glm with this formula in R Studio:

growthglm<-glm(FS_POLL_CONTRIB ~ PRECIP_fs +
                GPI_fs +
                PPI_fs +
                WEI_fs + PRECIP_fs:PPI_fs +
                PRECIP_fs:GPI_fs + PRECIP_fs:WEI_fs, family = gaussian(link="identity"))
summary(growth)

it give me this:

                   Estimate Std. Error t value Pr(>|t|)
(Intercept)      -169.80595         NA      NA       NA      
PRECIP_fs           0.26926         NA      NA       NA 
PPI_fs             24.27342         NA      NA       NA 
GPI_fs            -17.30260         NA      NA       NA 
WEI_fs             -8.59929         NA      NA       NA 
PRECIP_fs:PPI_fs   -0.03989         NA      NA       NA 
PRECIP_fs:GPI_fs    0.02980         NA      NA       NA 
PRECIP_fs:WEI_fs    0.01346         NA      NA       NA

How can I fix it so that the glm doesn't give me all the outputs as "NA" in R Studio? Or do you think this would be better with a generalized linear mixed-effects model (GLMM)?

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ as you mentioned "I only have one value of growth per each", so you get NAs for standard error.. there's no way to estimate from n=1 $\endgroup$
    – StupidWolf
    Commented May 17, 2020 at 19:26
  • $\begingroup$ What other test would you recommend me to do to evaluate it? $\endgroup$
    – Diego Centeno
    Commented May 17, 2020 at 20:45

0

Reset to default

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.