Lessons of Chemometric courses 7.5 hps

Introduction of Chemometric 7.5 hp- Models-DOE

Experiment Design experiment and SIMCA 12.0

Some definitions
Models required in Chemometric
- attempt to explain the systematic variation in a system
- expressed as mathematical equations
- statistical methods can be used to compare different models.
Measurements = model + noise
- include: Fundamental models(ex U = IR + c), Semi or empirical models (ex = b0 + b1x1 + b2x2 + b12x1x2 + c)
- Which model is considered to be a good model?
- clear and interpretable results, extracts ''maximum'' infor, lead to exact conclusions, give infor about new relating experiments.---> giving the precise predictions within the model limits.
The limitation of models
- Can not be correct, but some are superior than others.
- predictions can not be made without the model limits.
- the ''local model'' is to prefer instead of a '' global'' model.
- 2 variables are correlated, one does not cause the other.
The model
Y(reality, response variable) = (experimental factors-variables)X*(regression coefficients-variable weights)B + E(Residual-noise)

Model -or pertubation theory
Semi-empirical models: Taylor Series

Score plot: a summary of the relationships among the observations
Loading plot: a summary of the relationships among the variables, explain the pattern seen in the score plot.

2 plots are complementary and superimposable, directions in one plot corresponds to the same direction in the other.
DModX parameter indicates how well an observation fits the PCA model.
Design models: CCC, Box-Behnken( Sphere+center-points, Rotatable)
Lesson about the Principle component anaylysis(PCA)
Analysis of designed data and models 
Full factorial design (FF)
Fractional factorial design (FrF)
Old exams