2

I read about cochran's Q and don't think it is right for what I am trying to do. I created and example of my data.

I have 5 ecosystems, each of which had 3 sampling locations (plots). I wanted to compare the presence of floodwater between each ecosystem so a flood monitoring device was placed at each plot. The device recorded the flood daily as Presence/absence data.

I have attached a picture of how I plan on the data looking.

DATA

Any help on what stat would be appropriate would be much appreciated.

Douglas Putt
  • 31

1 Answers1

1

Fit a mixed effect logistic regression.

Data format:

     Eco    plot     day       Y
       1      1       1        1
       1      1       2        1
      .....

       5      15      6        0

Totally, there are 15*6 = 90 lines of data.

The model

$$\log\left(\frac {\Pr(Y=1)}{\Pr(Y=0)}\right) = \beta_0 + \beta_1 E_1 + \beta_2E_2 +\beta_3E_3 + \beta_4E_4 +\beta_5T + \gamma_s $$

where $E_i=I(\text{Eco} =i)$, T = time (1,2,...6), and $\gamma_s \sim N(0.\sigma^2)$ is the slot specific random intercept.

Maybe you need exclude the data from ECO = 1 and March? 15 (first day), because all of the $Y$s are y.

Of cause, you can modify the fixed effect part, for example, you can treat day as categorical, instead of continuous, you can add the interaction between day and Eco.

user158565
  • 7,461