Can I compare p-values?

#1
Hi all

I'm currently close to finishing my dissertation. I'm studying vegetational changes in quadrats (sample areas) inside and outside exclosures, to test for amounts of browsing damage from deer. I gathered my data over the winter (three study periods, November, December & January), so I want to know how much of the change in vegetation is from winter dieback.

To do this, I'm comparing change of height 'inside' November v December v January, 'outside' November v December v January and 'all quadrats' November v December v January with one-way ANOVAs. If 'inside' and 'all quadrats' have lower p-numbers, I am assuming that change was more significant in un-grazed areas and that therefore winter dieback is likely to have been an important factor in vegetation loss.

Here's an example:
In - p = 0.374
Out - p = 0.773
All - p = 0.456

Therefore, 'In' showed the most change, as there is a 37% chance that the null hypothesis that there is no significant difference between the means is true, the lowest chance of the three?

Is it possible to compare p-values in this way? I'm comparing a fixed number of sample points for each site. If not, can anyone recommend what test I should use in order to find out where change was greater?

Thanks in advance!
 
Last edited:

Karabiner

TS Contributor
#2
Your assumptions about p-values are wrong. You will have to test your hypothesis directely.

If you surveyed the very same quadrats each time, then oneway ANOVA for independent groups would be wrong.
Could you describe in more detail how your data were collected, and how large your sample size is?

With kind regards

Karabiner
 
#3
Your assumptions about p-values are wrong. You will have to test your hypothesis directely.

If you surveyed the very same quadrats each time, then oneway ANOVA for independent groups would be wrong.
Could you describe in more detail how your data were collected, and how large your sample size is?

With kind regards

Karabiner
Hi Karabiner,

I surveyed eight quadrats, four in and four out, at four sites over three months. Therefore I have 96 quadrats over all. Within each, I made two measurements, height and percentage cover, for each of nine categories of vegetation (Non-geophytic forbs, bryophytes, graminoids, geophytes, R. fruticosus, P. aquilinum, H. Helix, Woody Perennials, Bare Ground).

So for example, I want to compare how much change in height there was for 'bryophytes' in 'in' quadrats over the course of the survey, and I've used one-way ANOVAs to see if there was a significant change. In most cases, there wasn't (p ≤ 0.05). I was wanting to use the p numbers to compare different significances in difference over time.

I'm new to stats. What test should I use to do this if I can't use p values?

Thanks!
 

Karabiner

TS Contributor
#4
Seemingly, the surveyed quadrats are different between time points? If you would like to test whether the differences between time points are different between inside and outside, then you can use a two-way analysis of variance with time point (3 levels) and location (inside/outside). The interaction between these factors is what you are interested in.

If the quadrats were exactely the same at each time point, then you can use a repeated measures ANOVA with the additional “location “ factor and their interaction.

With kind regards

Karabiner