Im facing the condition where the R2 values is higher (>0.999) however the delta Q is greater.Thus, im not sure which one should i choose for the best kinetic model.Is anyone know and would like to share something, you are welcome to do so.TQSM.
There are many different measures of matching equations (models) to experimental data. They are based on a more or less complicated function of the difference between the calculated value from the equation and the experimental value. This can be the sum of the squared differences, the sum of modules of differences, or different weighted measures. There are no general rules regarding the choice of a given measure. It also does not seem wise to tie a given measure to an imaginary model order.
Usually I use the average relative error as the measure of fitting. As the objective function, I also use the average sum of squared differences. I rarely use R2, which usually works well with linear functions.
The measure "deltaQ" given by you makes no sense. It can have a positive value or maybe a negative one. It should be abs (xexp-xcal) / xexp. Obviously, the sum of such expressions divided by the number of experimental points is average relative error. Regards,
The most important in a kinetic model is the chemical reactions used to build such a model. Commonly, you need just to have a look at how close the theoretical and experimental curves
R2 greater than .999 is pretty impressive, I imagine both fits are pretty good. Could you share more information about what you are modelling, what the model will be used for and a plot of the data?
But I am more interested in the value of the average relative error. And not so much as the model, because I guess it, only the kinetic data. Because if it turns out that, for example, 90% of the data can be described by a constant function (upper limit), which happens in the kinetic studies of sorption processes (qt - indicates adsorption), R2 greater than 0.999 is the standard. Regards,
One thing you should is to plot the residuals (y-yhat) of the two models against all your explanatory variables included in the model (and also some that are not included) and see if there are trends in the plots. Trends will be an indication that the model structure is not very good enough.