Instructions for Covariate Correlation Analysis
- Basic Settings
- Correlation Method:
- Pearson: For linear relationships between continuous variables
- Spearman: For monotonic relationships, robust to outliers
- Kendall: For ordinal relationships, good for small sample sizes
- Correlation Threshold:
- Set minimum correlation value to highlight (0-1)
- Default 0.7 is commonly used
- Lower values (e.g., 0.5) for stricter multicollinearity control
- Higher values (e.g., 0.8) for more relaxed variable inclusion
- Non-numeric Variables:
- Option to exclude non-numeric covariates
- Categorical variables are automatically excluded if enabled
- Visualization Options
- Plot Type:
- Correlation matrix: Compact view of all correlations
- Scatter plot matrix: Detailed view of relationships
- Display Method:
- Color: Simple color-coded squares
- Circle/Square: Size indicates correlation strength
- Ellipse: Shape shows correlation direction and strength
- Shade: Uses color intensity
- Pie: Fills circles with pie charts
- Ordering Method:
- Original: Keep input order
- AOE/FPC: Order by eigenvectors or first principal component
- Hclust: Group similar variables
- Alphabet: Sort alphabetically
- Output Components
- Correlation Plot:
- Blue indicates positive correlations
- Red indicates negative correlations
- Color intensity shows correlation strength
- Values display exact correlation coefficients
- Highly Correlated Pairs:
- Table shows pairs exceeding threshold
- Lists exact correlation values
- Helps identify potential multicollinearity issues
Notes:
- Large correlations suggest redundant information between variables
- Consider removing one variable from highly correlated pairs before modeling
- Choice of which variable to remove should be based on ecological relevance and data quality
- Remember that correlation does not imply causation
- Missing values are handled pairwise in correlation calculations