Baseline Normalization of Time Frequency Power

Introduction

• Discussing baseline normalization
• Focus on choosing baseline time window and implications
• Baseline normalization shifts the y-axis without changing data time course

Importance of Interpreting Relative to Baseline

• Different baselines don’t change time course, only y-axis values
• Example with two different baselines illustrating interpretation change
• Interpretation of results changes as they become relative

Examples & Panel Illustration

• Example with three baseline periods
  • Baseline period 1: -500ms to -200ms
  • Baseline period 2: -300ms to 0
  • Baseline period 3: -100ms to +200ms
• Impact on post-stimulus activity and leakage

Temporal Leakage and Baseline Choice

• Temporal leakage effects
  • Activity leaking into pre-stimulus period
  • Baseline time window selection impacting results
• Best practice suggestion: baseline ending before time zero
  • Suitable periods: -500ms to -200ms, or -400ms to -200ms

MATLAB Implementation

• Demonstrating baseline time window selection in MATLAB
• Encouragement to experiment with different baseline time windows
• Differences in Percent Change and Decibel Normalization results

Baseline Normalization Across Conditions

• Different approaches: condition average vs. condition-specific baselines
• Example with three conditions (Red, Yellow, Blue)
• Condition-specific baseline preserves condition-specific power shifts
• Condition average baseline increases signal-to-noise ratio
• No single correct approach; context-dependent decision

Conclusion

• Importance of careful consideration for baseline selection and normalization