49 
Table D-1. Comparison of R-square and Root Mean Square for regressions to assist 
in determining the best root transformation with turbidity. 
root RSquare RootMSE 
1.1 
0.692513 
0.777389 
1.2 
0.694958 
0.774293 
1.3 
0.696466 
0.772377 
1.4 
0.697318 
0.771292 
1.5 
0.697708 
0.770795 
1.6* 
0.697773 
0.770712 
1.7 
0.697609 
0.770921 
1.8 
0.697286 
0.771333 
1.9 
0.696852 
0.771886 
2.0 
0.696342 
0.772534 
2.1 
0.695784 
0.773244 
2.2 
0.695196 
0.773991 
2.3 
0.694590 
0.774759 
2.4 
0.693978 
0.775535 
2.5 
0.693367 
0.776310 
2.6 
0.692761 
0.777076 
2.7 
0.692165 
0.777829 
2.8 
0.691581 
0.778567 
2.9 
0.691011 
0.779286 
*Highest r-square and lowest root mean square error are obtained for the 1.6 root of turbidity. This is very nearly 
matched by the results for the 1.5 root which was optimal for the VIMS data. Thus 1.5 root will be employed for 
further work. 
2. Does one K d -Turbidity model work for all tributaries? 
Analysis of covariance (ANCOVA) with an interaction term for Tributaries*turbidity 
was used to assess the consistency of the turbidity effect over tributaries (Table D-2). 
Table D-2. ANCOVA table showing test for consistency of turbidity (turb) effect over 
tributaries. "r1_5turb" is the rootl.5 transform of turbidity measurements. 
Source 
DF 
Type III SS 
Mean Square 
F Value 
Pr > F 
tributary 
16 
19.432 
1.214 
2.04 
0.0087 
rl_5turb 
1 
330.819 
330.819 
556.82 
<.0001 
rl_5turb*tributary 
16 
65.436 
4.089 
6.88 
<.0001 
appendix d 
Derivation of Regressions 
