The two major statistical analysis techniques used in this study were canonical correlation analysis and multiple linear regression. All statistical analysis were completed using the SAS Insititute's SAS 9.1 software.

Multiple Linear Regression

The first statistical technique used was a multiple regression analysis. This technique would be used to assess the feasibility of developing a predictive model that could be used to identify the spatial location of older age classes based solely on landscape features. Similar to simple linear regression, multiple regression analysis uses the principle of least-squares for estimating the regression parameters of predictor variables (Johnson 1986). Rather than a single response variable-predictor variable linear relationship characteristic of a simple linear regression, multiple regression requires multiple predictor variables for a single response variable.

In this analysis technique, stand age would be used as a response variable and distance to various water features, elevation, slope and aspect would be the predictor variables. A resulting regression equation would take the following form:

Stand Age = Y-intercept + B1(distance to water feature) + B2(elevation) + B3(slope) + B4(aspect)

Bx = regression parameters

Canonical Correlation Analysis

The second statistical technique used in the study is a canonical correlation analysis. A canonical correlation analysis is a common multivariate technique that can be used to investigate the relationships between two groups of variables that tend to naturally divide. This analysis technique searches for linear combinations of variables in each group in a manner that results in canonical variates that have the largest possible correlation (Manly 1986). The technique will produce a series of pairs of canonical variates, each representing an independent dimension of the relationship between the two groups of variables (Manly 1986).

In this investigation the two groups of variables used are species frequencies and landscape features (elevation, slope, aspect and distance to various water features). It is anticipated that the canonical correlation analysis would provide some insight into the relationship between these two groups of variables. For example, a desirable result would be multiple pairs of significant canonical variates with strong correlations. For the purpose of this example the first canonical variate pair (the pair with the highest correlation) could be represented by U and V and have a correlation of .83. U would largely represent stands with a considerable black spruce (1.15 x BS) and minor jack pine component (0.6 x JP) component while Y would represent stands with close proximity to large lakes (-0.9 x large lakes), minimal slope (-.98 x slope) and a north aspect (1.4 x aspect). This could possibly indicate that late succesional black spruce stands in the analysis landscape area are associated with large lakes, low elevations and north aspects.