...
Our (Ian Tse and I) first task was to organize the data and to identify which data corresponds to the matrices in the MathCad space. Then we try to deduce any trend that can be compared with the data. Our first guess was that these data can be represented by polynomial plots. From the error analysis, we determined that the third or fourth degree of polynomial suits the datasets the best. However, this polynomial regression method would only beneficial for fitting the experimental data but won't be useful for determining any actual parameters that representing the curve. Therefore, a saturation curve was suggested as presented here (the final figure). However, the regression didn't give a better fit for the datasets. We've tried to look from another perspective, and after several discussions, we decided that by normalizing the amount of settling (1-NTU/NTU~max~NTU ~max~), the analysis can be improved. The double-reciprocal analysis was imposed on the normalized datasets and we found that they conformed with Lineweaver-Burke analysis.
...