In Six Sigma, the DMAIC methodology seeks to improve a process at hand by analyzing its behavior in the past, and then, predicting a better performance in future. For this, the technique asks the 4 major questions of data analysis.
Description: What is the possibility of describing a collection of numbers in a meaningful summary?
To do a comprehensive study of the data, and to find a summarized meaning to it, the raw data needs to be meaningful. Statistical analysis, along with arithmetic calculations, can provide information that will then be used to forecast a better performance.
Probability: If a universe is known, can the outcomes be predicted? Deductive Logic seeks to answer such a question. A mathematical approach to find the possible outcomes of a given situation always leads to answers.
This method of data analysis is better than the enumeration approach, because it is more sophisticated and logical.
Inference: This deals with the prediction of an ‘unknown; universe, from the detailed knowledge and analysis of a ‘known’ universe.
This method of data analysis calls for inductive logic that cannot give the correct answers, but can only reveal the possibilities.
Homogeneity: If the data is coming from multiple universes, the statistical analysis predicts multiple outcomes. So, before proceeding to study the data in any way, testing for homogeneity is essential.
The baseline elucidates the details for starting the process improvements. In a certain example where 3, almost identical cases, are taken into consideration, how can homogeneity be defined?
It can be done by plotting the process behavior chart for each individual case. Then, by slicing the data and observing the X charts, the homogeneous nature of the systems can be understood.
With no homogeneity in the process, the predictive generalization for improvement becomes near to impossible. On the other hand, if the data is indeed homogeneous, it is possible to do a statistical analysis on the pile of data, and find probabilistic inferences to it.
Six Sigma practitioners use the ‘good judgment’ to organize data into subgroups. Tactics such as ‘rational sub grouping’ and ‘rational sampling’ prove to be efficient and prevent data from being placed randomly.
Therefore, a planned technique of analysis on a given set of data, proves more useful that randomly inspection.