Hello Lino,

You have a count variable (BioDiv) with two explanatory variables,

one categorical (GrRoof) and one ratio scale (Time).

Once you declare these as categorical and numeric respectively in your data object, a first cut model is an ANCOVA:

BioDivMod

Wow this is a great and very thorough explanation of a good possibility for my test, thank you! I am going to give the ANCOVA a try and see if the results make sense. Though, I am wondering why you chose to leave temperature out of that model? Since that is the ultimately the end determination of whether the green roof is performing well. We have learned about ANCOVA in my stats class, so I am a bit familiar with it. Would I just need to run the standard assumptions tests that are necessary for any ANCOVA?

Also, do I need to worry about autocorrelation with this ANCOVA model?

I am a bit skeptical of how I can really find out anything with only having looked at 6 green roofs, but that is really all I could manage to gain access too...

Thanks!

I was thinking, and I am not sure time really even matters here. To be blunt, should I even care about time or the date? Can't I just take the average temp for each green roof?

So my variables would just be:

green roof

number of plant species

avg temperature

I think this would be an ANCOVA, but a backwards ANCOVA, since instead of comparing two categoricals and controlling for a continuos (categorical ~ categorical + continuous), I would be comparing two continuous and controlling for a categorical (continuous ~ continuos +categorical)

so it would look like:

BioDivMod

Hello Lino

Your query was labelled "tests for biodiversity." And so in reply I took biodiversity to be the response variable. I overlooked your subseqent statement:

I want to find out, does increased biodiversity actually keep these roofs cooler?

For this question, temperature is the response variable. So the first step is for you to decide the response variable (y axis if you make a plot) and the explanatory variables (X axis on a plot).

If your question is about effects of biodiversity on temperature, you will need to state whether diversity was measure at each point in time. Or if only once, when.

Onward,

David S

Hi David,

I should have been more clear, but yes, I am ultimately interested in temperature. I forgot that the dependent variable is always on the left of the tilde.As for diversity, it was will just be measured at one time, with the assumption that the level of biodiversity has not changed.

Also, on another note, do you know how I would include time in such a model? The one other test I am trying to do is looking the surface temperatures of green roofs over a testing period, so like

lm(temperature ~ Green Roof + Time)

I think it is an lm I am trying to do. I tried doing ANOVA but did not know how to handle time. Do you know how I can model this? I have gotten different answers from different scholars, and am starting to think it really might be far more complicated than I had thought.

Hi Lino,

OK, so you are interested in the effects of biodiversity on temperature, where BioDiv is a numerical variable. The straight line model is

lm(Temperature ~ BioDiv).

You can check the straight line assumption by plotting Temperature against BioDiv. Equivalently, you can look at the residual vs fit plot from the regression of Temperature on BioDiv

Also, you seem to be interested in the effects of rooftype on temperature, where roof type is a categorical variable.

lm(Temperature ~ GrRoof)

And you have a time series, where Time is a regression variable (and a straight line looks unpromising(. The straight line model is

lm(Temperature ~ Time)

You can get more information by using BioDiv and GrRoof

Here is the model

lm(Temperature ~ BioDiv + GreenRoof+ BioDiv*GreenRoof).

Technically this is an ANCOVA. But we are going to ignore the label because ANCOVA has several uses, which confuses things.

Next. BioDiv and GreenRoof may well be correlated.

Which means that R (unless you specify otherwise) will use sequential rather than adjusted SS and MS and F ratios.

For sequential:

Mod1

The following paper will help you to find the required tests for green roofs to select appropriate growing medium:

Preprint Analysis of Physical and Chemical Properties of Alternative ...