Principal Circuit/Tear Advice
Analyses blocks to determine the principal circuit and potential tears to break us these blocks (see below)
Compute/Principal Circuit/Tear Advice (^T)
To analyze a block, select it by clicking in any of its squares and then press ^T. The results are displayed below:
The yellow squares indicate the principal circuit. Notice that the principal circuit may not involve every row in the block. The principal circuit is simply the longest circuit that the program has found within the block. (There can be several in fact; PSM attempts to find a good one.) In addition PSM builds a table of shunts. Shunts are paths from one node in the principal circuit to another node in the same principal circuit that does not touch the principal circuit expect where the shunt begins and ends. Shunts are used to construct some heuristics that assist the user in determining what relationships to weaken in order to break up the blocks.
Results/Tearing Advice Window:
Once a principal circuit is computed, the window below appears:
Down the left PSM shows the row names in the principal circuit (the length at the top) and some statistics about each row are generated. The statistics shown are interpreted as following:
A deeper discussion of these heuristic statistics is to be found in Donald V. Steward's book Systems Analysis and Management: Structure, Strategy and Design, which can be obtained from the author.
Generally one wishes to tear a node/item where WS is zero. Where there are several items with WS equal zero, generally tear the one of these where the values of S and F are the lowest. As one tears nodes then repartitions the matrix, look at these tears with an eye on whether they make good places the use assumptions. The program does not know this, but possibly you do. You will usually need to take a number of hacks at it before you get a set of tears with which you feel comfortable, or anyway least uncomfortable. The idea is that the program can give you the consequences of your various choices quite quickly so you can afford to try a number of choices.
When you start out you may find that you have to make many tears before you begin to see any emerging structure. Sometimes it is useful in this case and you have a reasonably long principal circuit to make multiple tears at the same time by double clicking on several nodes where the value of WS drops into a valley (with higher values on each side). But once having slashed through this brush, you can then start to try various alternatives.
(The weighting in the WS statistic is affected by the tearing decay value (Default 1.00) in Problem Maintenance Tab. Generally one is well advised not to play with it.)
How to indicate a tear:
There are two methods:
What to do after tearing:
Partition again to see the results of the tearing.
Generally, the simplest choice is to select the row indicated by the purple line. For this example if problem is repartitioned after tearing the purple lined row (promoting the marks to for example 2), the circuit is broken yielding smaller blocks as below: This provides a hierarchy of blocks, smaller blocks shown by different colors within larger blocks.
Eventually this process of repeating the Partition and Tearing process will render a matrix where no further blocks remain and the only marks above the diagonal represent the use of assumptions. Then the marks below the diagonal represent the ordering of the tasks showing the tasks must be done in sequence (unless you want to deal with more assumptions) or can be done in parallel.
You will then want to add tasks that are reviews of the assumptions. Then you an export to MS Project the basic set of tasks and their ordering. But you will have to add in the MS Project the reviews and explicitly show how you will handle the reviews and changes of assumptions.