Selasa, 12 Februari 2013

Glosarium ANP

  • NETWORK – A network is a structure of clusters and nodes and connections between the nodes that appears in a single window.   A network usually has feedback (lines going both ways between clusters) and may have inner dependence (a loop on a cluster indicating nodes in that cluster are connected to each other).  A hierarchy is a simple kind of network that has a goal node in a goal cluster that connects to criteria in a criteria cluster and so on, downwards, but no feedback.
  • MODEL – A decision model is a logical structure for making a decision.  Your decision model will have two files with a .mod extension.  The first is the main model that has the Benefits, Opportunities, Costs and Risks nodes, called the merits, in the top level network and a waterfall of subnetworks beneath.  The networks at the very bottom contain the decision alternatives.  This is the part of the model directly concerned the decision.  The second is the strategic criteria ratings model that is used to establish priorities for the B, O, C, and R in this particular decision.
  • NODES AND CLUSTERS – A cluster is a window inside a network that contains nodes.  A cluster is a logical grouping of factors or elements in the decision being considered.  The factors or elements are the nodes.  Clusters are a device to help you in structuring the decision problem.  All connections are made between nodes.  Clusters end up linked by default when a node in one cluster is linked to some nodes in another cluster.
  • CONNECTIONS – Connections are made by linking nodes together.  The way to decide on what connections to make is to pick a node and examine the nodes in another cluster (or its own) to see if they influence it.  If some of them do, link it to them.  This node will become the parent and the nodes it links to in a cluster will become its children for a pairwise comparison set.  A link will appear from the parent’s cluster to the children’s cluster.  Depending on how many clusters there are with children belonging to that node, a parent may have many comparison sets.  Inner dependence is when a node is linked to other nodes in its own cluster, and a loop will appear on the cluster.
  • BOCR – the Benefits, Opportunities, Costs and Risks nodes, referred to as the BOCR merits, are in the top level network in the main model. There is a goal node linked to them so at the end of the decision process you can insert their priorities.  To evaluate their importance one creates a separate strategic criteria model with ratings where the BOCR are the alternatives in the ratings spreadsheet.  You must know the most dominant alternative for the B, O, C, and R respectively as determined in your main decision model, and keep it in mind as you make the ratings.  For example, think of the impact of the dominant alternative for Benefits on the strategic criteria as you select the ratings across the Benefits row.  Finally the priorities derived in the ratings are inserted as direct data into the main model by selecting the Goal node and Compare/Misc./Data and typing them.
  • STRATEGIC CRITERIA – These are the invariant criteria or objectives of the individual or organization making the decision that always need to be satisfied and are external to the actual decision model of the day.
  • SUBNETWORK (OR SUBNET)  – a subnetwork is any network attached to a node in a network above it.  The BOCR nodes are in the top-level or main network.  Each of them has a subnetwork containing its control criteria nodes which will contain a hierarchical model of potential control criteria nodes.
  • CONTROL CRITERIA SUBNETWORK – there is a control criterion subnetwork for each of the BOCR merits.  In it one establishes a hierarchial structure of goal, criteria and perhaps subcriteria, pairwise compared to establish priorities.  These criteria or subcriteria are potential control criteria nodes. Select the nodes with the most priority to build decision subnets under (you should pick nodes with about 70% of the total priority).  To see the overall derived priorities, select Computations/Priorities and read the limiting priorities column.  These become the control criteria and you build decision subnets under them.  [If you have only criteria, their limiting priorities sum to 1.  If you have both criteria and sub-criteria you will need to multiply the limiting priorities by 2 to make them sum to 1 for the criteria and 1 for the sub-criteria, and determine the actual overall priorities (because of the way the limiting supermatrix “stochasticizes” its columns to make them sum to 1). 
  • DECISION SUBNET – the network attached to each control criterion node. You must create a decision subnet for each selected control criterion, so for benefits, for example, you may select 3 or 4 control criteria and end up with 3 or 4 decision subnets.  These are the bottom level networks, called decision subnets because they must contain a cluster named Alternatives that contains the nodes that are the decision alternatives.  In general, the other clusters and nodes you put into the decision subnet are the forces influencing the decision.
  • DECISION SUBNET TEMPLATE – it is often wise to build a general decision subnet containing the alternatives and all the other clusters and nodes you may need. You can build it in the first decision subnet in your model, and save that network, only, as a “template”.  You can then import the template by going into another decision subnet and opening the saved template.  It will overwrite whatever is there.  Modify as necessary by deleting clusters and nodes not needed, or adding new ones.  The original template will remain as it was and you can re-use it over and over again.  You may be able to make the node connections in the original template, but you will usually need to make the comparisons each time you  after importing as they will be different for different decision subnets.
  • FORMULAS – formulas are used to combine the results synthesized in the B, O, C and R subnets.  Formulas are only used in the top level model and can be selected, and changed, with the Design/Formulas command.  The first listed formula, the Additive (Negative) is automatically selected when you use the File/New command for a full model template.
  • SYNTHESIS – can be performed in any network.  The Computation/Synthesis  (or use the Syn icon) command always rolls up the results for all networks beneath the current network you are in when the command is invoked and combines them.  For the top level network the results from below are combined according to the selected formula.  Control criteria networks weight results rolling up from the decision subnets by multiplying by the the priorities of the respective control criteria nodes times the alternative values and adding them.
  • SENSITIVITY – can be launched using the Computation/Sensitivity command.  In a complex model, one usually wishes to do sensitivity for the B, O, C, and R nodes.  Look at the Tutorial file that is in Word on your CD to find out how to use synthesis.

Tidak ada komentar:

Posting Komentar