biology statistics - students t-test or other?

Question

I am doing a biology study and am not sure which statistical test to use to test for significance. I have collected data on tree ages and the number of fungal fruiting bodies which appear on them. I am trying to see if there is a correlation between age and the number of fruiting bodies i.e. the older the tree the more fungus. I think I should be using the t-test but am not sure any help appreciated -thanks

Answer 1

If you are doing a correlation then you could use the non-parametric test Spearman's rho (best for for ordinal data), or the parametric test Pearson's Product Moment Correlation for ratio/interval data. There are other criteria too - it would be a good idea to look in a stats book.

Answer 2

I wouldn't have thought a t-test is at all close to what you want. The amount of fungus is non-parametric data, the age is parametric if you're actually talking about age; but if you're talking about, how many seasons old/number of rings, then it's non-parametric too.

If you were going to split it up into "young trees" vs "Old trees", then I would have thought a Mann-Whitney AKA Wilcoxon rank sum test would have been what you wanted.

However, if you're doing tree age vs fruiting bodies... I've got no idea about non-parametric regression.

But it looks like Rozzy knows what she's talking about, so maybe go with her suggestion.

Answer 3

for biometry I definitely have continuously used Minitab. possibly you should attempt getting your palms on a duplicate of this software. blanketed with the utilizing is a statistical handbook that could help you take advantage of Minitab for undertaking statistical checks.

Answer 4

T-test is fine. A t-test is really the same as testing the beta coefficient of a simple regression, where you regress fungus amount ("Y" variable) on tree age ("x").

Update: You are free to try non-parametric tests here, but it is seriously not necessary, and in addition parametric tests always are more robust than non-parametric tests if you can do them. Don't complicate things by splitting into groups and then doing group comparison (I don't see the point of that). You can collect very useful statistical data even when the values are only 0,1 and STILL do correlations/regressions.

The two suggestions below I find highly confusing. Just do the simple regression and see if your beta coefficient is significant. It's easy and quick.