I am conducting a biological research about species abundance variations between different hours and localitions.

I am comparing 11 different locations and 5 different hours.

Independent variables are hour and locality.

Dependent variables are abundances of each species found in the locations.

Each hour treatment has 11 distinct samples.

Likewise, each locality treatment has 5 distinct samples.

Kruskal Wallis test was chosen for comparing each abundance with hour and location variables, because two factor ANOVa requires assumptions that dont fulfill my data.

Some species are present at certain hour of the day but are absent at another hours, that means there will be some zeros in my data, corresponding to species absent.

For that reason i eliminate from statistical testing hours or treatments with zeros in all of their samples. But.

What sould i do if in a 11 samples treatment:

1. I have 7 samples equal to zero...

2. I have 8 samples equal to zero...

3. I have 9 samples equal to zero...

4. I have 10 samples equal to zero...

On the other hand...

Some species are present at certain locations but absent at another location, that means there will be some zeros in my data, corresponding to species absent.

For that reason i eliminate from statistical testing treatments with zeros in all of their samples. But.

What sould i do if in a 5 samples treatment:

1. I have 4 samples equal to zero...

2. 3 samples equal to zero...

3. 2 samples equal to zero...

4. one sample equal to zero...

Thanks for reading.