I'm currently doing a bit of biology coursework in which i have to test species diversity against depth of soil on a hill. My teacher says that I cannot use Spearman's rank to calculate the correlation between these two variables, as it is a stats test and somehow not compatible. The way i see it, if i can plot a graph of these two variables, then surely I can find the correlation between them quantitatively? I don't know whether it is his opinion or Edexcel's, but to me it seems wrong. Could anyone help me?

Well, Spearman's Rho is for ordinal (non parametric) data. If you want to run correlations on parametric data (in this case number of species against depth of soil, both ratio level of measure) you should use Pearson't product moment correlation coefficient.

 

However, your teacher's objection appears to be that it is not appropriate because it is a statistical test. I'm not sure why this might be, possibly s/he is considering other factors involved?

 

Basically, you can test for correlations between any two variables, whether or not it's actually sensible to do so (e.g. IQ and shoe size). Occasionally, by chance, you will come up with a significant relationship that means absolutely nothing. This is where many people go wrong and is one of the sources of spurious statistics.

 

As you (or your teacher) know more about your project than I do, you would have to ask yourself 'is this a sensible question'? and 'Would the answer mean anything?'.

I have used PMCC for other data in the coursework. The exam board wants us to use spearmans rank in the coursework, but according to my teacher I cannot use it for the measures my hypothesis focuses on, and have to do species frequency against height. It seems I have to guess the correlation between sdi and height. It seems totally wrong to me, but he's telling me to 'correct' it.

 

The other thing I can't understand is that im trying to prove the assumption that the height of the soil on the hill has an effect on its water content. I don't know any other way to do this other than correlate the two variables, but apparently this too is wrong, as I am comparing two environmental factors????

 

Thanks for your help

Well, if you want to show an effect of one on the other (i.e. height on the hill on soil water content), correlation isn't the way to go as you can never assume a causal relationship from a correlation. Although a correlation between the two would support the hypothesis that the higher on the hill you are, the less water content you are likely to find. You just couldn't say it was because you were higher on the hill.

 

If your teacher wants you to do species frequency by height, that suggests you will have categories for each of the different species you are counting. In this case, you will need to break height into gategories also, and then use a Chi Squared test. You can't use Spearman's Rho on catagorical (nominal) data.

 

What is your hypothesis by the way?

Hypothesis:

There is a relationship between the depth of soil on a pingo on foulden common and the species diversity supported by it.

 

(A pingo is a small hill)

 

The profile of the pingo was divided into 8 height zones, the occurences of each species recorded at each one of them, and then sdi values calculated. 9 soil samples were taken to estimate the mean water content of the soil at each height zone.

Well the 8 height zones can be ranked in a meaningful way. What is sdi? something like species diversity index or something? If that gives you continuous (scale) data, then you can use a correlation to test your hypothesis.

I meant SDI to be species diversity index, we use Simpson's reciprocal index at A2. I think the values are scale data, since i can plot a graph of them?

I have no idea. It's not my area. I heard "We use blah blah blah at A2". It doesn't mean anything to me and the A2 is a road to Dover. I can't help you with that side of things.

Fair enough, thanks for your help anyways

No worries. Find out what level of measurement the SDI yeilds, and also exactly what your teacher has agains you using correlation.