Basic Modelling with AgenaRisk
Contents
1. asena Basic Modelling with AgenaRisk www agenarisk com Copyright Agena Limited All rights reserved Copyright Notice 2013 Agena Ltd No part of this publication may be reproduced published stored in an electronic database or transmitted in any form or by any means electronic mechanical recording or otherwise for any purpose without the prior written permission of Agena Ltd THERE IS NO WARRANTY OF ANY KIND FOR THE ACCURACY OR USEFULNESS OF THIS INFORMATION EXCEPT AS REQUIRED BY APPLICABLE LAW OR EXPRESSLY AGREED IN WRITING Contents Copyright Notice 1 2 2 1 22 2 3 2 4 2 5 2 6 2 7 3 1 3 2 4 Introducing the Car Costs Model Building the Car Costs Model Creating Ranked Nodes Using Weighted Expressions Using Partitioned Expressions Customising Risk Graphs Adding Constants Using Arithmetic Expressions Completing the Model Running the Car Costs Model Creating and Deleting Scenarios Adding Observations to Scenarios Next Steps 22 22 23 26 Modelling with AgenaRisk 1 Introducing the Car Costs Model This tutorial takes you through the process of building a realistic predictive model using AgenaRisk In the process of building this model you will be introduced to some of the more advanced and powerful features that AgenaRisk has to offer such as ranked nodes simulation nodes partitioned expressions and continuous graphs You will also gain modelling experience that yo2. 3 Bring up the properties dialog for the new node Change the name to Miles per Gallon change the unique identifier to mpg and click Apply 4 In the Appearance tab change the background colour of the node to light blue and change its shape to being a rounded rectangle 5 Click Apply 6 Now click the Node Probability Table tab on the left of the dialog This initially displays the Normal expression that was automatically created when created this simulation node 7 In the NPT Editing Mode drop down box select Partitioned Expression Partitioned expressions are very powerful They allow us to generate an NPT using a different expression for each combination of parent states The dialog looks like Figure 10 In this case we can specify a different distribution of miles per gallon for a small car for a medium car and for a large car This accurately reflects the real world one distribution for miles per gallon wouldn t be sufficient to capture these differences a Node Probability Table NPT Editing Mode Partitioned Expression w Node Details Select the required parents from the list on the left and add them to the list on the right The list on the right will contain the parents involved in the partitioned table The order of the parents determines the configuration of states in the table below Node States a j lt lt Remove all lt Remove Node Constants Enter a formula for each part
3. Click on the Scenario 1 text You will see that it is editable Using the Backspace key remove the text and type Mercedes Right click anywhere in the Scenario Manager and choose Add a New Scenario Rename this scenario Honda Add another new scenario and name this one Ford Click on the Risk Table tab below the Scenario Manager Notice that the column of fields is called Mercedes This corresponds to the first scenario you added By default only the first scenario in a model is marked as active In order to see the other two scenarios on the Risk Table click their Active boxes You should now see all three scenarios 22 Modelling with AgenaRisk Risk Scenarios GGG Active Display on Risk Graphs Mercedes rj v Honda wj wj Ford wj wj Risk Map Risk Table Mercedes Honda Ford 2 Car Costs Car type No Answer v No Answer v No Answer v y Reliability No Answer v No Answer v No Answer v Maintainability No Answer v No Answer v No Answer v Miles per Gallon fuel_price_const miles_per_year_const Price per mile Annual Fuel Cost Annual Maintenance Cost Total Annual Cost Figure 23 Scenario Manager showing three scenarios 10 Scenarios can also be deleted from the Scenario Manager Add two new scenarios and then select both of them by holding down the Ctrl key between clicks 11 Right click on the scenarios and select Delete Selected Scenario s Verify that the Scenario Manager now contains just the three scenarios
4. For the expression in the Medium cell specify 28 as the Mean and 50 as the Variance For the expression in the Large cell specify 18 as the Mean and 30 as the Variance This partitioned expression now represents our view that larger cars are less fuel efficient than smaller cars The quickest way of completing multiple cells with a very similar formula is to use copy and paste Right click on a completed cell and select Copy then right click into all the other cells and click Paste Each cell is now identical Then double click on each cell to modify it 12 13 14 15 16 Click OK Clear all evidence in the model and then run a calculation by clicking the calculate button on the toolbar Double click on the Miles per Gallon node Recall that you only specified one state for this node You can now see how simulation has created a number of states automatically Make the node larger so you can see all the states Hold your mouse over some of the bars on the graph and notice the information that is available This information is also available on the Risk Graph Panel Look at the window on the Risk Graph Panel marked Miles per Gallon Click on the Expand State List hyperlink You will now see a full list of the states and probabilities that exist on this node You will need to scroll down to see all of them If you are displaying Risk Graphs on the Risk Graph Panel rather than on the Risk Map you can still acces
5. Click on the insert button Insert Type fuel_price_const in the Constant Name field and type 3 in the Default Value field Click Apply Click on the Node Probability Table tab and then in the Expression Type drop down box select Arithmetic Delete the 0 0 value in the Arithmetic Expression text box Right click on the Arithmetic Expression text box select Insert constant fuel_price_const and click OK You have now made Fuel price gallon a constant node This is because the node is a simulation node and its NPT is generated from the value of the fuel_price_const constant This means that after calculation Fuel price gallon will have a single value This value will either be the default 3 or whatever is specified in the corresponding Risk Table entry Before the value of the fuel_price_const constant can be modified an entry needs to be added to the Risk Table Click on the Risk Table tab right click on the Fuel price gallon entry and select Properties In the Connection Type drop down box select Constant and then click OK The value of fuel_price_const can now be changed by editing the contents of this entry s text box Return to the Risk Map copy the Fuel price gallon node paste it and move the copy below the original Open the properties dialog for the newly pasted node change its name to Miles per Year and change its unique identifier to mpy Click on Apply Click on the Node States tab Ch
6. Type node after editing Modelling with AgenaRisk 12 Copy the Car Type node and paste the copy to the right of Car Type on the Risk Map 13 Change the properties as follows Unique Identifier Reliability High gt Low 3 leave the states as High Medium and Low Background Colour Light Blue Table 1 Properties for Reliability node 14 Your two nodes should now look like this Car type Figure 3 Risk Map after adding Reliability node 15 Copy the Reliability node and paste it on to the Risk Map beneath the two existing nodes and change the properties as follows Unique Identifier Maintainability Background Colour Medium Blue Table 2 Properties for Maintainability node 16 Draw edges from Car Type to Maintainability and from Reliability to Maintainability This captures the information that Maintainability is influenced by both Car Type and Reliability 17 Your Risk Map should now look like the one shown in Figure 4 below Modelling with AgenaRisk Car type Figure 4 Risk Map after adding Maintainability node and edges 2 2 Using Weighted Expressions 1 Currently the model shows the causal relationships between the three nodes Car Type Reliability and Maintainability but for these relationships to be of any use you need to define the Node Probability Table NPT of the Maintainability node 2 Right click on the Maintainability node bring up the properties dialog and click on the Node Pro
7. created earlier If you have a scenario containing a lot of observations the quickest way of creating another scenario that is only slightly different is to right click on the scenario and select Clone Selected Scenario s This will create a copy of the scenario that can then be renamed and modified 3 2 Adding Observations to Scenarios 1 Observations can be added to scenarios either via the Risk Map or the Risk Table The only exception to this is values for constants they can only be entered via the Risk Table You will enter the constant observations first For all three makes of car fuel price and annual mileage are going to be the same In the row entitled fuel_price_const enter the value 3 in each scenario s text box Likewise enter 10000 in all three boxes in the miles_per_year_const row Your Risk Table should now look like this Mercedes Honda Ford Car Costs Car type No Answer v No Answer v No Answer v Reliability No Answer v No Answer v No Answer v Maintainability No Answer v No Answer v No Answer v Miles per Gallon fuel_price_const 3 3 3 miles_per_year_const 10000 10000 10000 Price per mile Annual Fuel Cost Annual Maintenance Cost Total Annual Cost Figure 24 Risk Table with constant values entered for three scenarios Now enter the Car Type for each scenario Select Large in the drop down box for Mercedes select Medium for Honda and select Small for Ford Similarly enter the Reliabil
8. 0 Click on Apply If another dialog appears then click OK on it this is a benign error message that occurs because we haven t yet updated the node probability table for this node Click on the Node Constants tab Right click on the constant and choose Delete Selected You don t need any constants defined on this node Click on the Node Probability Table tab The expression used for generating the NPT of Price per Mile is going to be a simple division of Fuel price gallon by Miles per Gallon don t worry about dividing by zero AgenaRisk handles this internally Clear the Arithmetic Expression text box and then right click inside it Select Insert parent Fuel price gallon fuel_price Then type the division symbol Finally select Insert parent Miles per Gallon mpg Click on the Graph Defaults tab Click on the Graph type for large data set drop down box and select Line Uncheck Treat Min Max X as percentiles checkbox and clear the Min X and Max X boxes Click OK Run a calculation Double click on the Price per Mile node to bring up its Risk Graph and resize it so that you can see it all The graph should look like the one shown in Figure 15 14 Modelling with AgenaRisk Figure 15 Risk Graph for Price per Mile with long tail 13 There is a very long tail on this distribution that isn t of interest You can modify the Risk Graph properties for just this node so that the tail isn t d
9. 0 a lower bound of 0 and an upper bound of 600 7 When Maintainability is Low use a TNormal with a mean of 500 a variance of 150 a lower bound of 0 and an upper bound of 600 8 Runacalculation 9 Double click on Annual Maintenance Cost resize the Risk Graph and verify that it looks like the one in Figure 19 18 Modelling with AgenaRisk Figure 19 Risk Graph for Annual Maintenance Cost 10 Add a simulation node below Maintainability on the Risk Map Make Annual Maintenance Cost and Annual Fuel Cost the parents of this new node by adding edges between them 11 Open the properties dialog for the new node and supply the following property values Unique Identifier total_ annual cost Node States 10 10 inf Arithmetic Annual Maintenance Cost amc Annual NPT Expression Fuel Cost afc Background Colour Medium Blue Lower percentile X Bound Upper percentile X Bound Table 6 Properties for Total Annual Cost node 12 Run a calculation 13 Double click on Total Annual Cost resize the Risk Graph and verify that it looks like the one in Figure 20 19 Modelling with AgenaRisk Figure 20 Risk Graph for Total Annual Cost 14 The model is now complete The Risk Map should look something like this Car type Figure 21 Completed Risk Map of Car Costs model 15 Save your model In the next section you will interact with the model and use it to generat
10. Advanced Modelling with AgenaRisk and explore the range of well documented example models that come with the tool 26
11. Total Annual Cost node and compare the three makes of car Move your mouse over the lines to view the summary information Given the information we have entered so far the model tells us that owning a Mercedes will have the highest annual cost mean 2309 followed by Honda mean 1346 Ford is the cheapest mean 1076 The narrower a distribution the more confident we are that the mean represents the true value of the node Notice that the distribution for Ford is the narrowest we can be fairly confident that the actual annual cost of owning a Ford will be close to 1076 Conversely the Mercedes distribution shows the greatest spread the actual annual cost of owning a Mercedes could be significantly higher or lower than 2309 Look at the two nodes Car Type and Reliability Remember that we entered observations for these nodes in the Risk Table Each scenario is represented by a label sitting on top of the node In the case of Reliability notice that the label for the Mercedes scenario says Soft Evidence and that this scenario has two corresponding bars on the Risk Graph The two bars show the information that you entered namely that a Mercedes could just as likely have high or medium reliability Imagine now that having read a glowing review in a car magazine you believe the reliability of Mercedes cars to be high Right click on the Reliability node and select Enter Observation Mercedes High Run a calculat
12. at appears type Car Costs and click OK Create a new node in the top left corner of the Risk Map Change its name to Car Type change its type to Ranked change its unique identifier to car_type and then click Apply Your dialog should look like the one shown in Figure 1 below 10 11 Modelling with AgenaRisk _ Node Details Node Mame Node Probability Table Node Constants Figure 1 Node Details for Car Type node Changing the unique identifier to something short yet meaningful is important because later on this node will be used in expressions Click on the Node States Tab In the Type field in the middle of the dialog choose Low gt High 3 Then in the States text box replace the values Low Medium and High with Small Medium and Large Click Apply and then click on the Appearance tab Select the rounded rectangle at the top of the dialog change the background colour to dark blue and click Apply The colour scheme is arbitrary but in this model we are going to use dark blue for nodes where the user has to make a choice medium blue for nodes relating to cost and light blue for nodes containing assumptions Colour coding nodes can make models more user friendly Click on the Text Format tab Change the font size to 12 change the text colour to white and then click OK Your node should now look like this with a red exclamation mark showing that the node needs to be edited Car type Figure 2 Car
13. bability Table tab 3 Rather than defining the NPT manually we are going to make life simpler and use an expression that will behind the scenes generate the NPT automatically In the NPT Editing Mode drop down box select the value Expression The expression editor will appear in the dialog as shown in Figure 5 Node Probability Table NPT Editing Mode Expression v Expression parameters take the form of standard mathematical expressions and can include node names available by right clicking in the parameters text field Node States If a parameter is badly formed the text field will have a red border You can find out the problem by holding the mouse over the field 2an Expression Type Node Probability Table 4 TNormal Node Constants Mean Ed Variance Notes B Appearance Text Format Lower Bound Upper Bound 7 Reset fields to default values l v Set current field values as new defaults Cancel Figure 5 Expression editor for Maintainability node 4 Click on the Expression Type drop down box and observe that only two types of expression are available for ranked nodes TNormal and Arithmetic We will deal with Arithmetic expressions later For now leave the type set to TNormal Modelling with AgenaRisk 5 TNormal is short for Truncated Normal This is a Normal distribution where the two ends are cut off or truncated When the TNormal distribution is used with Ranked
14. cker to edit the variance and mean text boxes by hand 12 13 14 15 16 Click on Apply This generates the underlying NPT You can inspect the NPT that has been created by selecting the value Manual in the NPT Editing Mode drop down box Scroll around the table and look at the values in each of the cells When you are finished set the value of the NPT Editing Mode drop down box to Expression Click OK Now you can test this model fragment Click on the calculate button on the toolbar o Notice that the red icon has now disappeared from the Risk Object in the Risk Explorer and the exclamation mark has also disappeared from the nodes themselves This means that a calculation has been executed and that the model is in a consistent correct state Modelling with AgenaRisk 17 Double click on all three nodes on the Risk Map so that their Risk Graphs appear Resize the nodes so that you can see the graphs in their entirety Your Risk Map should like the one in Figure 9 Car type Reliability Figure 9 Car Type Reliability and Maintainability after calculation 18 The Maintainability Risk Graph tells us that when we know nothing at all about Reliability or Car Type it is most likely that Maintainability will be Medium 19 Try entering different observations for both Car Type and Reliability and notice how the probability distribution of Maintainability changes Verify in particular that when Car Type is Small a
15. e predictions about the expected running costs of your three short listed makes of Car 16 Before proceeding to the next section you might wish to check your solution against the correct solution which can be found here Examples Tutorials Car Costs ast 20 Modelling with AgenaRisk 21 Modelling with AgenaRisk 3 Running the Car Costs Model 3 1 Creating and Deleting Scenarios 1 One of the most powerful features of AgenaRisk is the ability to compare different scenarios side by side You are going to use scenarios to compare the three car makes on your short list Mercedes Honda and Ford Before continuing turn off the auto calculate feature by toggling the auto calculate button a You will be adding and removing a lot of observations in this section of the tutorial and you dont want a calculation to occur automatically after each change Click on the downward arrow at the top of the screen next to Risk Scenarios x This expands what is known as the Scenario Manager Currently the Scenario Manager contains only one Scenario as shown in Figure 22 below al Risk Scenarios Active Display on Risk Graphs Scenario 1 Pj wj Figure 22 Scenario Manager showing one scenario When a new model is created it has just one scenario This scenario is active which means it is displayed in the Risk Table and that observations in this scenario are shown on the Risk Map In addition the scenario is displayed on Risk Graphs
16. eck the Make upper bound positive infinity box There are now two states Change the lower bound of the first state to 0 Change the lower bound of the second state to 1 The two states should now be 0 1 and 1 inf Click on Apply Click on the Node Constants tab change the value of the Constant Name field to miles_per_year_const and change the Default Value to 10000 Click on Apply and then in the Node Probability Table tab verify that the expression now contains the modified constant name miles_per_year_const In order to refresh the underlying node probability table click on the red hyperlinked message if it appears 13 Modelling with AgenaRisk 21 Click OK 22 Look in the Risk Table and notice that a new entry has been created for miles _per_year_const Car Costs Car Type No Answer r Reliability No Answer x Maintainability No Answer r Miles per Gallon fuel_price_const miles_per_year_const 23 Figure 14 Risk Table showing entries for constants Save your model 2 6 Using Arithmetic Expressions 1 2 10 11 12 Copy the Fuel price gallon node paste the copy and place it to the right of the original Draw edges from Fuel price gallon and Miles per Gallon to this node Open the properties dialog for the copy change its name to Price per Mile and change its unique identifier to ppm Click Apply Click on the Node States tab Change the lower bound of the single state to
17. he states of the parent nodes 21 Runa calculation 22 Double click on Annual Fuel Cost and resize the Risk Graph so you can see all of it It should look like this 16 Modelling with AgenaRisk Figure 17 Risk Graph for Annual Fuel Cost Remember that you can get more screen space by decreasing the width of the Risk Graph panel and Risk Explorer or by retracting them completely 23 Your Risk Map should now look something like the one in Figure 18 Car type Figure 18 Partially complete Car Costs Risk Map 24 Save your model 2 7 Completing the Model 1 There are now just two nodes left to add then the Car Costs model will be complete and you can start using it and viewing its predictions 17 Modelling with AgenaRisk 2 Add a simulation node to the right of Maintainability Make Maintainability the parent of this new node by adding an edge between them 3 Open the properties dialog for the new node and supply the following property values Lower percentile X Upper percentile X Bound Table 5 Properties for Annual Maintenance Cost node 4 The NPT is specified using a Partitioned Expression Specify a TNormal expression for each state of the parent Maintainability 5 When Maintainability is High use a TNormal with a mean of 100 a variance of 100 a lower bound of 0 and an upper bound of 600 6 When Maintainability is Medium use a TNormal with a mean of 200 a variance of 15
18. ion Observe how the mean total annual cost of owning a Mercedes has decreased slightly to 2156 Despite the improved reliability however owning a Mercedes is still likely to cost more than owning either of the other two cars Return to the Risk Table view and try entering different values for Fuel price gallon and Miles per Year Remember to run a calculation when you have finished making your changes Return to the Risk Map to view the updated Risk Graphs or expand the Risk Explorer so that you can see the graphs while you are still in the Risk Table view Note what effects the changes have had on the Total Annual Cost node Try entering different constant values for each scenario For example you might decide that if you owned a Ford you would drive it a lot more than if you owned a Honda and that if you owned a Honda you might drive it slightly more than you would drive a Mercedes You would represent this information by specifying different values for miles _per_year_const 18000 for Ford 10000 for Honda and 8000 for Mercedes say Under these circumstances a Ford costs the most to run 25 Modelling with AgenaRisk 4 Next Steps This tutorial has shown you step by step how to build a reasonably complex model using AgenaRisk There are however still many ways of using the tool that have not been covered In order to learn more about these different uses you should you should work through the more advanced tutorial
19. isplayed This allows you to focus more clearly on the main area of the distribution 14 Bring up the properties for the Price per Mile node Select the Graph Defaults tab 15 Mark the checkbox next to Treat Min Max X as percentiles 16 Type 0 1 in the Lower percentile X Bound box and type 98 in the Upper percentile X Bound box This configures the graph so that the first 0 1 and the last 2 of probability mass are not shown 17 You should see a graph that looks like the one in Figure 16 below So cS LA io ht ht Ca Ca 7 L Ca Laj Figure 16 Risk Graph for Price per Mile after changing graph properties 18 If you have time enter different values of Car Type and observe the effects on Price per Mile Before continuing clear all observations 15 Modelling with AgenaRisk 19 Create a new simulation node below and to the right of Price per Mile Draw edges from Price per Mile and Miles per Year to this new node 20 Open the properties dialog for the new node and supply the following property values using the techniques you have previously learned Arithmetic Price per Mile ppm Miles per Year mpy Background Colour Medium Blue Table 4 Properties for Annual Fuel Cost node Upper percentile X Bound es Specifying the states for simulation nodes should not matter because Li AgenaRisk automatically determines the correct ranges for nodes conditional on the evidence and t
20. ition by double dicking the cell ee Car type Small Medium Large qes 5 Notes B Appearance Reset fields to default values vw Set current field values as new defaults Cancel Figure 10 Partitioned expression for miles per gallon 8 Select Car Type in the left hand list and click the Add gt button to move it to the right hand list This is how we indicate that we want to supply different expressions for each state of the parent node Car Type Notice that there are now three cells in the table below the two lists one for each state of Car Type 9 Double click on the first cell An expression entry dialog appears In the Expression Type field select TNormal Fill out the parameters so that the dialog looks like that shown in Figure 11 Modelling with AgenaRisk Expression parameters take the form of standard mathematical expressions and can include node names available by right clicking in the parameters text field If a parameter is badly formed the text field will have a red border You can find out the problem by holding the mouse over the field Expression Type TNormal Figure 11 Partitioned expression entry dialog 10 Click OK Notice that the TNormal expression now appears in the first cell in the table 11 Complete the other two cells in the same way with TNormal expressions For both expressions specify the Lower Bound to be 5 and the Upper Bound to be 100
21. ity for each make of car For Honda select High and for Ford select Low For Mercedes you are going to enter soft evidence Double click on the Reliability entry to expand it Enter 1 in the Medium and Low text boxes 23 Modelling with AgenaRisk To enter soft evidence on any other scenario you need to make any other Soft evidence can only be entered on the first active scenario in the Risk Table scenarios temporarily inactive 7 Your Risk Table should now look like the one in Figure 25 Car Costs Mercedes Honda Ford Car type Large v Medium v Small X Reliability Soft Evidence w High v Low lt Car Costs Reliability Observation New Node dpo Hig rj 1 0 Medium fal ho tow Maintainability No Answer v No Answer v No Answer X Miles per Gallon fuel_price_const 3 3 3 miles_per_year_const 10000 10000 10000 Price per mile Annual Fuel Cost Annual Maintenance Cost Total Annual Cost Figure 25 Risk Table with constants hard evidence and soft evidence 8 Run a calculation 9 When the calculation is complete switch to the Risk Map view Your Risk Map should look like this Car type 100 100 100 honda Medium H ford LS ford Small Figure 26 Risk Map showing three scenarios 24 10 11 12 13 14 15 16 17 Modelling with AgenaRisk Observe that each Risk Graph now shows all three scenarios Look at the
22. nd Reliability is Low Maintainability is skewed towards High This illustrates the increased weighting you gave to Car Type 20 You should also look at the effects of so called back propagation Clear any observation you may have entered on Car Type or Reliability and enter the observation High for Maintainability Note that the other two distributions both change Car Type moves towards Small and Reliability moves towards High but Car Type is more skewed In other words if all we Know about a car is that its Maintainability is high then the most likely explanation is that the car is small 21 Now enter the observation Large for Car Type The distribution for Reliability changes dramatically we are now almost certain the Reliability must be high 22 Before continuing select Tools Clear Entered Data All and click OK in the confirmation dialog 23 Now save your model 2 3 Using Partitioned Expressions 1 Create a new simulation node beneath Car Type on the Risk Map Move the existing nodes to create space as required A simulation node can be created by clicking on the button shown below on the toolbar C Press OK in the dialog informing you about simulation node defaults By making the node 8 Modelling with AgenaRisk a simulation node you are telling AgenaRisk to dynamically calculate the optimum set of states that the node should have 2 Draw an edge from Car Type to the node you just created
23. nodes it is always truncated using the values 0 and 1 This is because Ranked nodes are all mapped on to an underlying scale of 0 1 6 You are now going to create a weighted expression that defines the parameters of the TNormal distribution This weighted expression will reflect the relative strength of influence that the two nodes Car Type and Reliability have on Maintainability 7 Right click anywhere in the Mean text box and choose Insert weight expression WeightedMean 8 You will see a dialog like the one shown in Figure 6 Weight Elicitation Certainty Uncertainty Please select the required parents _ Car type _ Reliability Figure 6 Dialog for eliciting weighted expression 9 This dialog allows you to select which parents you want to include in the weighted expression and what weighting you want to give them The slider bar at the top lets you specify how much certainty you have about this weighting Complete the dialog so that it looks like the one in Figure 7 In this case the expression says that when defining the maintainability of a car we are moderately certain that car type is twice as important as reliability Weight Elicitation Certainty Uncertainty Please select the required parents Car type A Reliability Figure 7 Weighted expression dialog for Maintainability node 10 Click OK The expression editor should now look like Figure 8 11 Modelling with AgenaRi
24. og 11 Modelling with AgenaRisk 11 Bring up the Miles per Gallon Risk Graph You can now see a vertical dotted line indicating where the mean sits in the distribution Figure 13 Risk Graph for Miles per Gallon node showing mean The mean will displayed every time you reopen the Risk Graph on this node but the mean will not be displayed for any other nodes 12 Enter different values for Car Type and notice how the Miles per Gallon Risk Graph changes 13 Clear all observations and save the model 14 If you have time try modifying the other Risk Graph properties and experiment with the hierarchical property system Consult the user manual for more information about any properties that you don t understand 2 5 Adding Constants 1 Add a simulation node to the Risk Map beneath Miles per Gallon and bring up its properties dialog 2 Modify the following properties Background Colour Light Blue Table 3 Properties for Fuel price gallon node 12 10 11 12 13 14 15 16 17 18 19 20 Modelling with AgenaRisk Give the node one state with a lower bound of 0 5 and an upper bound of 10 Click on the Node Constants tab Constants are numbers to which you can assign names so that they can subsequently be used in expressions The power of constants Is that their values can be modified later on via the Risk Table without needing to edit the expressions in which they are involved
25. s the information summary by dragging the 10 Modelling with AgenaRisk separator bar on the far left of the graph to the right so that the window splits 2 4 Customising Risk Graphs 1 2 Right click on the Miles per Gallon node and select Properties from the pop up menu Select Graph Defaults from the left side tab Click on Graph Type for large data set drop down box select Line Uncheck Treat Min Max X as percentiles checkbox and clear the Min X and Max X boxes Click OK on both dialogs Figure 12 Continuous Risk Graph for Miles per Gallon node To view the details of any point plotted on the graph just move your mouse over it The Risk Graph properties system is complex but powerful It works hierarchically You specify a set of defaults at the node level These will apply in all cases except for those scenarios in which you choose to override them Then for a given scenario you can change the way its Risk Graph is plotted by selecting the Scenario Properties tab in Properties for a node Now let s assume that you want the Miles per Gallon node and only this node to have its mean shown whenever its Risk Graph is displayed Right click on the Miles per Gallon node select Properties and click on the Graph Defaults tab Check the Mean box in the Summary Statistics section of the dialog This ensures that the mean will now always be shown on the Risk Graph for this node 10 Click OK on the properties dial
26. sk D Node Probability Table NPT Editing Mode Expression v Node Details Expression parameters take the form of standard mathematical expressions and can include node names available by right clicking in the parameters text field pa Node States If a parameter is badly formed the text field will have a red border You can find out the problem by holding the mouse over the field Expression Type gt 4 A N TNormal y Mean wmean 2 0 car_type 1 0 Reliability Node Constants ee Variance Notes 6 Appearance Text Format Lower Bound Upper Bound Reset fields to default values v Set current field values as new defaults Cancel Figure 8 Expression editor after defining a weighted expression You can see that the level of certainty you specified using the slider bar in the weighted expression dialog is used as the variance for the TNormal distribution The mean of the TNormal is a formula based upon your choice of parent nodes and their weights Note that the unique identifiers of the parent nodes are used in the formula rather than their names If you want to edit this expression in future e g if you want to change the weights or variance you can right click on the Mean text box and select Insert weight expression WeightedMean The existing values will appear in the weighted expression dialog and you can modify them However you may find it qui
27. u can later apply to your own problems Before working through this tutorial you should read the tutorial Getting Started with AgenaRisk to ensure that you understand how to use the common features of AgenaRisk Assume you want to predict the running costs of a new car You have short listed three different makes of car Mercedes Honda and Ford You want to find out which of these is going to be cheapest to run over time The important variables under consideration are e Car type small medium or large e Reliability which impacts on maintenance cost e Fuel economy miles per gallon e Fuel price e Annual mileage Modelling with AgenaRisk 2 Building the Car Costs Model 2 1 Creating Ranked Nodes 1 Start AgenaRisk and create a new model by selecting Create New Model from either the opening dialog or the File menu Alternatively DI you can click on the new model icon on the toolbar If it is not already selected toggle the auto calculate button on the toolbar 2 Ensure that the Risk Explorer is expanded and that the item called New Risk Object is selected There will be only one Risk Object in this model Notice that the Risk Object is flagged with a red icon HD You will see this icon in various parts of the AgenaRisk interface It means that a calculation must be run on the Risk Object in question in order to make it consistent Right click on the New Risk Object item and select Rename In the dialog th
Download Pdf Manuals
Related Search