Origins
Chemometrics The application of statistics and mathematical methods of chemistry. Recognized as a branch of Analytical Chemistry. Basic methods were originally developed in the elds of economics and psychology.
!
These elds are commonly faced with complex, interrelated data sets. Examples. Prediction of economic trends based on various indicators. Measurement of intelligence.
Why Chemometrics?
Most analytical procedures attempt to make a problem univariate. Look at only a single unknown material. Mask the presence of other materials. Remove potential interference. Hold all experimental conditions constant except for analyte.
Why Chemometrics?
Some problems are complex by nature. The composition of many components contribute to a materials overall properties. Some instruments produce a huge number of measurements per sample.
GC/MS example.
A typical run may produce 2000 spectra with a mass range of 500 m/e. This results in 1,000,000 data points for a single sample. Too much information to take in at once.
response
Chemometrics
Chemometrics has been considered a branch of analytical chemistry since the mid 1970s. The introduction of personal computers is one key factor for its increased use.
time
in spe fo c rm tr at al io n
Over 12,000 references that contained the word or concept chemometrics were found during a recent SciFinder search.
Chemometrics
It is not a single tool but a range of methods including: Basic Statistics Signal Processing Method Optimization Factor Analysis Factorial Design Resolution Detection Pattern Recognition Library Searching Neural Networks
Chemometrics
New or modied approaches are introduced very rapidly. We will attempt to cover most procedures that have found signicant use in analytical chemistry.
Converting data to knowledge
What are data?
Dened as things known or assumed from facts and gures from conclusions. Broadly dened, data are simply raw information - both qualitative and quantitative. Raw data are meaningless. We need some form of analysis and model to gain knowledge.
Data model
analysis
Information
Knowledge
Kinds of data
Soft labels, descriptors, category assignments (qualitative) - the water is hot Hard numerical (quantitative) - the water is 400 K Avoid numbers that are based on soft assignments.
Types of data
Natural
Arise from natural phenomena. These are factors we cant control. Example: it was raining that day or the temperature was 94 oF when we did a study.
Experimental
Measurement of a property under known, controlled conditions. These are laboratory conditions.
Commonly, both types are involved in a study.
Types of data
Types of data
Discrete data
Only a nite possible range and interval is possible.
Classify each as continuous vs. discrete, hard vs. soft and natural vs. experimental. Score on a multiple choice exam pH of Lake Erie Weather conditions Color of a ower Current time
Continuous
Occur over a range but are not discrete. Limits of an instrument may give the appearance of being discrete
Obtaining meaningful data
If your data is bad, nothing can save it. To collect good data, you must have a plan The rst step should be to ask a few questions.
What is the desired outcome? What is the population? What are the parameters? What do we already know or can assume? What is the basic nature of the problem? research, monitoring, conformance ...
Other factors to consider
Temporal nature of the problem
Long range, short range, one-time
Spacial nature
Global, limited area, local
Other related factors
Most are common sense. However, it requires that you look at the entire problem before starting any work.
Tools well use
When dealing with large data sets, modern computer equipment and software is a great asset. Primary software will be MS Excel (with the data analysis add-on) and XLStat (another Excel add-on). Other tools will be introduced when needed.
