Finding patterns in data

What is Pattern Analysis?

PATN is a software package that performs Pattern Analysis. PATN aims to try and display patterns in complex data. Complex in PATN's terms, means that you have at least 6 objects that you want to know something about and a suite of more than 4 variables that describe those objects. Data must be in the form of a spreadsheet of rows (the objects in PATN) and the columns (variables), as in Microsoft Excel™.

There are usually around 7 components to a 'realistic' (read as adequate, comprehensive, fair, reasonable or intelligent) pattern analysis in PATN-

  1. Import the data
  2. Check the data using PATN's Visible Statistics functions. It is important that you are confidant that your data is error-free. PATN will operate on what data is available. There have been examples where conclusions have been drawn from flawed data. Don't be among them!
  3. Possible data transformation or standardization. Make sure the data are in a form where the association, classification and ordination will will make have greatest opportunity to detect patterns.
  4. Generate association values between objects and / or variables. Association in PATN  (resemblance, distance, dissimilarity, affinity etc) is a quantitative estimate of the relationship between each pair of objects and / or variables. This is probably the most important step in PATN. Understand how the different options work!
  5. Classify. PATN has two options to organise the objects and / or the variables into a set of discrete groups. If there really are well-defined groups in the data, PATN will easily find them. In most cases, there is a gradation between groups, so objects may be marginal. User-defined groups or groups from other applications can be imported and analysed. Think of classification as reducing n-objects to k-groups. Makes information easier to view!
  6. Ordinate. This is the most powerful technique in PATN. Think of ordination as reducing the number of significant variables to 2 or more usually, 3. With 3 new variables, visualization of the objects (not variables in PATN) is easy!
  7. Analyse the results. This is where you come in! PATN can do most of the computational work, but the hard work is in interacting with PATN to interpret the results. Name the groups from a  classification! Detect the trends from an ordination!

PATN is setup to make it easy for you to follow this process. For the average dataset, the first 6 steps should take no more than 5 minutes! The 7th step could however take many hours as you come to grips with what PATN is trying to say about your data. The 3-dimensional plot in PATN is the most powerful component for helping you with step 7.