Related Experiment Videos
Note on the use of correlations in speech timing
The Journal of the Acoustical Society of America
|December 1, 1986
Summary
Researchers using Pearson's correlation for speech analysis may encounter statistical artifacts from overlapping intervals. This study demonstrates this artifact and corrects flawed reasoning in prior replies.
Area of Science:
- Speech Science
- Acoustic Phonetics
- Statistical Analysis
Background:
- Pearson's product-moment correlation is widely used in speech research.
- Analysis of temporal organization in speech often involves correlating overlapping intervals.
Purpose of the Study:
- To demonstrate that high correlation coefficients can be a statistical artifact when analyzing overlapping intervals in speech.
- To mathematically refute incorrect reasoning presented in a previous reply regarding this statistical artifact.
Main Methods:
- Mathematical analysis of correlation coefficients derived from overlapping intervals.
- Critique of statistical reasoning in Tuller and Kelso's reply (J. Acoust. Soc. Am. 76, 1030-1036 (1984)).
Main Results:
- High correlation values can arise spuriously due to the method of using overlapping intervals.
- Identified and mathematically corrected flawed arguments concerning this statistical artifact.
Conclusions:
- The method of correlating overlapping intervals in speech analysis can produce misleading results.
- Researchers should be cautious of potential statistical artifacts when interpreting correlation coefficients in temporal speech organization studies.