Critical Study of Approximating Functions (and Methods) as Applied to Ocean Station Data. Project 1. Regression Analysis
MISSOURI UNIV-ROLLA DEPT OF STATISTICS AND APPLIED MATHEMATICS
Pagination or Media Count:
The stepwise multiple regression technique is used in a model building process to develop predictors of temperature, salinity, and sound velocity as functions of geographical location, time, and depth. Models which give reasonable results are obtained through successive trials using higher order terms of the independent variables. The model for sound velocity yields values which are nearly identical to the Wilson sound velocities contained in the ocean station file and values computed using a modified version of the MacKenzie equation. The distribution of residuals resulting from comparisons of the Wilson equation sound velocities to those obtained from the regression model both computed from actual temperature and salinities shows that 98 fall within the range of plus or minus 2 msec. A comparison of the regression model sound velocity values computed from regression predictions of temperature and salinity with the Wilson values shows that 88 of the residuals fall in the range of plus or minus 12 msec. The results, which are valid for the 4 degrees square centered at 37.5 degrees North latitude and 69.5 degrees West longitude, are discussed in terms of the statistical significance of the distribution of the residuals. Since the physical characteristics of the area selected are rather complex, the application of this technique to other parts of the ocean is recommended.