Looking for guidance on an experimental design

#1
Hey folks:

I've got a question that requires a somewhat complex experimental design (nested + repeated measures) and wanted to get a few opinions.

I am measuring soil respiration (essentially metabolic rate of microorganisms and plants underground) in a forest that has been damaged by a tornado, and an undamaged control.

I have 12 plots (6 damaged, 6 undamaged). but respiration has a lot of spatial variation, so I am taking at least 10 measurements within each plot. Here is the second complication. Respiration depends mostly on temperature and moisture, so I expect it to vary seasonally, increasing in warm months, and decreasing in cool months. Therefore, I will remeasure soil respiration each month 10 times in each of the 12 plots

So my question is: Should I pick 10 random spots within each plot to subsample over and over during that time, or would it be better to pick 10 new random spots each time? The goal of the analysis is to measure the difference in respiration between damaged and undamaged forest and see if this difference persists over time.

Thanks for any advise you can offer
 

noetsi

Fortran must die
#2
I dont understand why (if you think the variation in temperature is between months) you are measuring multiple times in one month to address this. It would seem the way to address this would be to measure in different months and control for this. I may misunderstand your point or method :)

You want to measure what you are controlling for or think the key variables are. I would think if month is a key variable and spartial variation then you need samples that allow you to compare by month and by spartial variation. For example if soil ph and temperature were both issues you would want to have say an acid and non acid sample in month x and the same in month x+2.

On the other hand I had a single graduate DOE course so that might be simplistic.
 

Dason

Ambassador to the humans
#3
Does the act of measuring soil respiration have an effect on future measurements of soil respiration? Or are you assuming that the act of taking the measurement doesn't change anything? I don't know how you measure soil respiration so this might be obvious to you but I have no idea.
 
#4
noetsi: I am not measuring multiple times in one month. I am measuring points within each plot once per month (to account for spatial variation; these are nested and will end up averaged together essentially). Essentially I have 12 plots and 10 points nested within each plot (120 total).

My question is: Should I return to the same 10 points within each plot each month to remeasure, or pick 10 new points each month.

Dason: measuring soil respiration is non-invasive and measuring it in one month shouldn't effect it the next month.
 

noetsi

Fortran must die
#5
I may simply not understand, but if you measuring ten points within each plot it would seem to me you are measuing multiple times in one month. Regardless of that, if you want to compare across months, and I think you would given your comments on temperature, then you have to control for factors other than months. The simplest way to do that is to measure the same points over different months. That way they will be the same except for month. Alternately you can do designs like block designs that allow multiple factors beside month (really temperature and moisture level) to be considered.
 
#6
i think u'd need plenty of samples collected at different temp and moisture. If u will use the same location than incase respiration is affected by previous periods', than you might like to check the partial auto correlation. Then u input the correlated periods, temp, moisture to a nonlinear regression system and set the target as respiration. There are mixture designs for DOEs might help u.