1
$\begingroup$

I have the following question concerning ecological statistics in a scientific study I am performing. I have measured the amount of copepods (little invertebrates) at different locations and at different times of the year together with other parameters from the water. Thus:

  • I have one single responsible variable y (in this case the number of live copepods in $1m^3$ of water).

  • I have multiple predictor variables $x_1$ to $x_i$ (in this case temperature, acidity, salinity, amount of ammonium in the water, phosphate).

  • I have done this at five different locations and for ten different months of the year each time. For a total of 50 different y variables measured and 250 x variables measured.

My hypothesis is that the copepods do not react to any of these environmental factors and they thrive independently of temperature, acidity, etc.

I thus would like a test to demonstrate the lack of correlation or lack of prediction power of the x variables. But I'm not sure what to do. How would you proceed? How would you express the certainty that these x do not influence y?

$\endgroup$
1
  • $\begingroup$ What is the typical number of live copepods in 1m3 of water? How are you counting them? $\endgroup$ Oct 13, 2018 at 3:56

1 Answer 1

1
$\begingroup$

Fit a mixed effect Poisson regression model. Fixed: $x_1$ to $x_i$ and month; random: location specific intercept. Of course, the response variable $Y$ is the number of live copepods in 1m$^3$ of water.

Statistical test is used to demonstrate that response variable $Y$ and covariate $x_i$ has relation. Here you want to see they has NO relation. Then presenting the confidence interval (CI) is better than presenting p value.

$\endgroup$

Your Answer

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

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