|
I have been pondering this for some time. I
am running a simulation experiment where I take 1000 sample plots each from
point patterns of different properties. I am trying to summarize the
results from the experiment using K-analysis. I am interested in the mean
value of K across all plots, but I am also interested in the number of plots
that were significantly aggregated or significantly regular out of the 1000
points. My question is this, what is the proper way to construct the Monte
Carlo intervals? Do I have to do it at the plot level (i.e., run 1000
simulations for each plot--this would be ALOT of simulations) or could I use a
group (experimentwise) interval?
If I use a group approach, I run into an additional
problem. For example, if I construct them traditionally say with an
alpha=0.05, I could run 1000 simulations (ksim) of CSR and then used the ranked
values ksim[25] and ksim[975] at each distance d to give me the bounds.
But, I would then expect 5%, or 50 of my plots, at chance to be significantly
aggregated or regular. So would it be appropriate to Bonferronize the
p-value of the interval; i.e., calculate a Monte Carlo interval at a p of
alpha/number of plots = 0.00005?
Just wondering, because this REALLY influences
my conclusions in the experiment.
Mike
|
* By using the ai-geostats mailing list you agree to follow its rules ( see http://www.ai-geostats.org/help_ai-geostats.htm )
* To unsubscribe to ai-geostats, send the following in the subject or in the body (plain text format) of an email message to [EMAIL PROTECTED] Signoff ai-geostats
