I need some help to decide which experimental design should I consider to analyze an experiment with Mahogany (Swietenia macrophylla) regeneration tree.

The main objective is to investigate the density of mahogany regeneration around the mother-tree in Amazonia forest. Was selected 20 trees (not randomized trees) and from each tree four oriented-plots (North, South, East, West) was instaled. Each plot with 5 meters x 40 meters long. Into each plot the number of Mahogany regeneration was counted at sub-plot 5 meter away of each other (5-10; 15-20; 25-30; 35-40).

The idea is to investigate if theres a specific orientation where occur more regeneration and until wich distance they reach.

To consider the variation between trees, four Blocks was created based on the DBH of each tree (investigate if DBH has an influence on the experiment). Each Block has four trees (repetition). Each tree has 16 treatments (four orientation x four distance class). The idea of Blocks is to control the great variation in dimention of the trees (DBH from 70 to 140 cm). I did not tried DBH as a co-variate!

I run a Poisson and Negative Binomial model in Glimmix of SAS (I hade to sum 1 to each value because SAS was did not converge with zero values!)

PROC GLIMMIX DATA=MAHOGANY;

TITLE 'POISSON DISTRIBUTION';

CLASS BLOCK ORIENT DIST;

MODEL NREG= ORIENT|DIST / DIST=POISSON LINK=LOG;

RANDOM BLOCK;

RUN;

PROC GLIMMIX DATA=MAHOGANY;

TITLE 'NEGATIVE BINOMIAL DISTRIBUTION';

CLASS BLOCK ORIENT DIST;

MODEL NREG=ORIENT|DIST / DIST=NB LINK=LOG;

RANDOM BLOCK;

RUN;

However, I think this design is a Split-plot because Factor B (Distance) is nested into Factor A (Orientation). However SAS did not converge the Split-plot statistical model!

There is some suggestion for analyzes?

Thiago