Help with model selection
Thank you for your reply, Bob! MONOVA was one of the options on my list, till I realized that it actually compares one or more categorical independent variable(s) with two or more treatment levels AND more than one continuous response variable (chemicals in my case). I don't know where to accommodate my explanatory variables (i.e., soil nutrients) in a MANOVA. As far as I understand it, MANOVA is more suitable when you have discrete groups, I have both. All of my response and explanatory variables are continuous variables. Would you recommend adonis PERMANOVA, ANOSIM or MRPP? I did book an appointment with the UoA statistics department, thank you for pointing to it. I really appreciate the time you?ve taken out of your schedule to help me out. -Simon
On Fri, Mar 16, 2018 at 3:04 AM, Bob OHara <bob.ohara at ntnu.no> wrote:
The design is good. But the details will matter. I think the best advice I can offer is to seek advice! Your university has a consulting service: http://www.stat.ualberta.ca/~tcc/#home, and I think half an hour with them looking at your data with you will help a lot. You have a multivariate response, so a MANOVA is probably needed. The full (horrible) model you have is Time*Treatment*Moisture/Site/Tree where Time is before/after, and Treatment is the inoculation. Hopefully you can remove some of the interactions: the Time*Treatment*Moisture terms are important, so you would hope that the interactions with Tree and Site don't matter. Tree should probably be a random effect, although it is less important her because your design is good. I think giving more advice than this would require some poking around the data, but I hope this is a help. Bob On 03/16/2018 03:33 AM, Altaf Hussain wrote:
Hello everyone,
I have some field collected data from *90* random trees on *6* sites and
I am comparing the trees for the concentrations of different chemical
compounds. The chemicals are of two different types, *(Group 1 = A, B, C,
D, E, F and their sum, X)* & *(Group 2 = G, H, I, J and their sum, Y)*. I
also have the sum of X and Y that I call Z. All variables are continuous
and use the same units. The 6 sites are closely located but are unique in
terms of their soil nutrients (N, P K, S and their sum, Tot.Nutrients,
*continuous/same
units*) and soil moisture. I further grouped the sites for soil moisture
and I now have 2 Dry, 2 Medium dry and 2 Wet sites, (6/3=2), character
variables. Each site has 15 trees and an equal number of trees i.e., (5)
were treated using different fungal inoculation densities i.e., (0 for
control, 4 and 16). The trees were sampled in a pre and post inoculation
manner, where the pre inoculation samples serve to determine the baseline
chemistry of trees and the post inoculation samples will show the chemical
responses of the trees to inoculations *(if any!).*
Alright, first thing first, I admit that I am very new to statistical
modeling, which is why I have such a complicated design, and which is why
I
am posting ?. The following are the questions that I want to answer using
my dataset introduced above:
1) How do I compare my 6 sites for trees' chemical responses to the
different inoculation densities?
2) How do I check if any of the nutrients, their interaction, or their
interaction with soil moisture is playing a role in the chemical
variation?
I am open to any suggestions that will improve my study!
I am using RStudio version 3.4.3 for the analysis.
I appreciate any input that will help me with the analysis.
Regards,
//Simon
[[alternative HTML version deleted]]
_______________________________________________ R-sig-ecology mailing list R-sig-ecology at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
-- Bob O'Hara Institutt for matematiske fag NTNU 7491 Trondheim Norway Mobile: +49 1515 888 5440 Journal of Negative Results - EEB: www.jnr-eeb.org
_______________________________________________ R-sig-ecology mailing list R-sig-ecology at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology