English user manual
Contents
1. 436971 422200 978842 992200 50569 67900 102408 101030 207088 230600 850189 810000 86958 105100 Relative Error RE The model was validated and is considered as accurate enough _ Aggregation Statistics MMRE 0 11 Median 0 00 Skewness 0 50 Pred 25 0 90 lQRange 0 18 Kurtosis 2 11 Copyright Fraunhofer IESE 2011 Usage of COBRIX Validation of the CoBRA Model The second tab Productivity shows the productivity estimation improvement in figures as well as in a diagram Here improvement is understood as the reduction of the derivation of calculated nominal productivity The dark blue bars in the diagram are the actual productivity values of the historic projects based on their cost and size The dark red dots are the cor responding calculated nominal productivity values The magenta line is the estimated overall nominal productivity based on the project data of all his toric projects Validation of CoBRA Model Reale Nominal Productivity o 8 o amp E i 8 8 a g 5 z o eri Stability Statistics normalized Std Derivation Actual 0 23 Improvment actual nominal 4 97 normalized Std Derivation Nominal 0 05 Done The third tab P Details contains a table of the relevant size cost and productivity data for each project the table can be easily copied to Excel2. 0 chooses the standard visualization Re mark The visualization of the causal model has not been officially released yet lt entry key PRODUCTIVITY_MODEL gt 0 lt entry gt C 2 Predefined Scale Types The answer possibilities on the question sheets that are needed for the elici tation of past and new projects are defined by the scale type of the respec tive question The predefined scale types for new CoBRA projects are de fined by the following properties in the configuration file Remark Scale types can be added modified and removed in each CoBRA project inde pendently For example lt entry key SCALE_TYPES3 gt Frequency lt entry gt lt entry key SCALE_TYPE3_APO gt Most of the Time lt entry gt lt entry key SCALE_ TYPE3 AP1 gt Occasionally lt entry gt lt entry key SCALE_ TYPE3 AP2 gt Infrequently lt entry gt lt entry key SCALE_TYPE3_AP3 gt Rarely lt entry gt Copyright Fraunhofer IESE 2011 40 Configuration File Predefined Cost Factor Categories C 3 Predefined Cost Factor Categories The cost factors can be categorized by default into Personal Product Process and Project With the help of the property below you can add additional categories or modify existing ones For example to add an additional category insert lt entry key CF_CATEGORY5 gt Customer lt entry gt Warning The change of categories has an effect on all COBRA projects C 4 Change DB Server S
3. A 1 JAVA Runtime Environment Installation If you have not installed Java in version 6 or higher or if you are not sure which version of Java is installed on your computer please go to www java com and follow the installation instructions you find there A 2 COBRIX Installation The installation of the CoBRA Tool itself is very simple Just unzip the folder CoBRIX_R2 5 2 in the CoBRIX_R2 5 2 zip file to a direc tory on your hard disk e g C Program Files CoBRIX Usage of COBRIX Starting COBRIX CoBRA Project Selection Dialog B Usage of COBRIX In this section we provide a short introduction on how the CoBRA Tool can be used to collect data build and validate an estimation model and how to use it for cost estimation risk analysis and benchmarking We start by ex plaining how a new CoBRA project can be created and how an old one can be reloaded B 1 Starting COBRIX CoBRA Project Selection Dialog To run the application double click on the file COBRIX exe which can be found on your hard disk in the COBRIX_R2 5 2 folder When you start the CoBRA Tool a dialog similar to the one below appears CoBRA Project Selection Seles You can choose to start a new CoBRA project i e create a new model by pressing the New button or continue working with an existing one by choos ing the appropriate project and pressing Open or you can import an XML file containing a COBRA model project Import
4. COBRI Release 2 6 User Manual Michael Klas 27 06 2011 Table of Contents The Installation JAVA Runtime Environment Installation COBRIX Installation Usage of COBRIX Starting COBRIX CoBRA Project Selection Dialog Import CoBRA Project Initialization of a New CoBRA Project CoBRA Process Overview Export COBRA Project Import Cost Factors synthesize Organization Relevant Cost Factors Export Cost Factors Expert Management Cost Factor Ranking by Experts Choose Most Important Cost Factors Build Causal Model Collect Multiplier Data synthesize Multiplier Data Choose Past Project Data Type Elicitation of Past Project Validation of the COBRA Model Enter New Project Data Project Simulation Estimate Project Cost Analyze Project Cost Risk Benchmark Projects Configuration File Causal Model Visualization Predefined Scale Types Predefined Cost Factor Categories Change DB Server Settings Debug Settings Productivity Model Project Simulation Other Properties OOONDARWN ND A The Installation The following installation description assumes that you are using a Microsoft Windows PC If you are using another operating system the procedure may be slightly different The installation of the CoBRA Tool CoBRIX is very simple but to execute CoBRIX you must have installed the newest ver sion of the Java Runtime Environment or at least version 6 or higher which is available from the website www java com
5. e f you chose to work with an existing project the project overview dialog appears showing the actual state of the CoBRA project see CoBRA Process Overview Copyright Fraunhofer IESE 2011 Usage of COBRIX Import CoBRA Project e If you chose to create a new CoBRA project the first step is the project initialization see Initialization of a New CoBRA Project e If you chose to import an XML file containing a COBRA model you can select the XML in a file dialog see Import COBRA Project and then re turn to the project selection dialog B 2 Import CoBRA Project By pressing the Import button you get to the following dialog P EIB tS Filename Initial_20070918 cobra xml My Network Places Files of type COBRA Project XML cobraxmli In this dialog you can import an XML file cobra xml with an existing Co BRA project model After choosing a file and pressing the Open button you can choose a name for the imported CoBRA project model After a valid name has been chosen you return to the CoBRA Project Selection Dialog Usage of COBRIX Initialization of aNew CoBRA Project B 3 Initialization of a New CoBRA Project Initialization of your CoBRA Project CoBRA Project Name My CoBRA Project Context automotive industry embedded C maintenance Project size metric Project cost metric Uni P Definition kilo lines of code without comments person hours Expert expe
6. n a ol in oOo I 0 EIN nm con BE The dialog shows all experts who have provided estimations for the chosen cost factor variable Moreover it shows the experience of the experts and the divergence of their estimation from the estimations of the other experts 25 Usage of COBRIX Choose Past Project Data Type In the last column you can remove the check to avoid the use of the outlier estimation in the cost overhead calculation E10 Lars Ritter B 15 Choose Past Project Data Type By pressing the Past Project Data Type button you get to the following dialog Select Past Project Data Type DE Past project data are available in the form of size cost AND cost overhead estimation data strongly recommended With the check you can choose to use or not to use cost overhead estima tion data of past projects Not using cost overhead data usually results in strong overestimation of project costs Therefore we strongly recommend keeping the checkmark B 16 Elicitation of Past Project By pressing the Elicitate Past Project button you get to the following dialog 2 Projects for Elicitation Name Project Fda Project Tokio Project Fdd Copyright Fraunhofer IESE 2011 26 Usage of COBRIX Elicitation of Past Project In this dialog you can manage our past project database You can add addi tional past projects modify existing ones or remove them from the data base
7. factors must have either a positive or a negative influence Variables that connect to their factors do not need an influence e The field Why does the cost factor have this influence can provide ad ditional information for the model user B 12 7 Causal Model Validation The graph has to fulfill a set of criteria rules in order to be compliant The Check Li function validates the graph As soon as the graph is validated and no warnings fa appear the Accept button is enabled in the tool bar e The graph contains at least one loop e The graph contains two factors variables with identical names e The graph does not contain exactly one dependent variable Effort e The dependent variable Effort has one or more outgoing influences e The dependent variable Effort has no incoming influences e A factor is not connected to the dependent variable or to an influence that targets the dependent variable e A variable does not have exactly one outgoing influence e A variable contains one or more sub variables e A variable has an indirect influence e A variable has no scale type selected e An influence contains more than one indirect influence 21 Usage of COBRIX Collect Multiplier Data B 13 Collect Multiplier Data By pressing the Collect Multiplier Data button you get to the following dia log Expert s Multiplier Data Collection BE Data Collected Collect The
8. you get to the following dialog Benchmark Project Project Name Project IESE G1 Probability Distribution Cummulativ Probability in 208 253 CostOverhead in Mean of the Project Cost Overhead o 250 3 Lower Quantil of Cost Overhead Mean of all Past Projects 182 Upper Guantil of Cost Overhead Mean of all Past Projects Copyright Fraunhofer IESE 2011 Usage of COBRIX Benchmark Projects This dialog shows us how difficult in terms of cost overhead our new project is It is compared to the earlier projects past projects and can be signifi cantly easier than the average if it is in the green area left area or signifi cantly harder if it is in the red area right area It may also be of average difficulty if it is located between the upper and lower quantile boundaries yellow area 39 Configuration File Causal Model Visualization C Configuration File This chapter explains how to use the config xml file to customize the tool behavior for one s own special needs The configuration file can be found in the installation directory of the CoBRA Tool e g C Program Files COBRIX Generally the settings provided by the default configuration cover most ap plication scenarios Therefore only experienced users should change the configuration file C 1 Causal Model Visualization The property allows choosing the type of visualization for the presentation of the causal model A value of
9. 2011 Usage of COBRIX Cost Factor Ranking by Experts Ranking of Cost Factors by Reinhart fe Personal Process Product Project Cost Factor PROC Disciplined requirement management Ae Bahasa gt 0 01 PROC 3 Implementation of reviews and inspections PR OC 4 Disciplined quality management PROCS Quality of testing ah More Important W Less Important Definition Extent of user custamer participation whether the users are providing information reviewing requirements documents performing some of the analyses themselves and taking part in acceptance testing In this dialog the experts can rank the cost factors for each category To do this they use the More Important and Less Important buttons d More Important gr Less Important which move the selected cost factor in the list one position up more im portant or down less important To support the ranking process the defini tion of the selected cost factor is shown To navigate through the individual categories a ss Personal Process Product Project the Back and Next buttons can be used After the cost factors of all categories have been ranked you can close the dialog and store the rankings with the Finish button 11 Usage of COBRIX Choose Most Important Cost Factors B 11 Choose Most Important Cost Factors By pressing the Choose most important button you get to the following dia log C
10. 42 The tool allows choosing the productivity model that is applied to transform calculated cost overhead into project cost A value of 0 chooses the stand ard model lt entry key PRODUCTIVITY_ MODEL gt 0 lt entry gt If this property is set to 1 the productivity model presented in the first Co BRA paper Lionel et al 1998 is used i e cost 1 b1 co b0 size If this property is set to 2 the productivity model developed together with the Tool is used i e cost b1 co 100 bO size If this property is set to 3 the productivity model developed for Web CoBRA is used i e cost b1 co 100 size If this property is set to 4 the previous productivity model is used i e cost 61 co 100 size but b1 is calculated as the median of inverse produc tivity to increase robustness If this property is set to 5 the productivity model does not consider the in fluence of cost factors i e cost b1 size Copyright Fraunhofer IESE 2011 Configuration File Project Simulation If this property is set to 7 a productivity model with a definable fix cost val ue is used i e cost b1 co 100 size fix_cost where the fix costs can be defined by an additional property lt entry key FIX_COST gt 0 lt entry gt C 7 Project Simulation These settings influence the project simulation in the Simulate Project step which is responsible for samplin
11. Message Box Section In the message box section you can see all relevant information warnings and exceptions of the current graph e Simple activity information is preceded by the icon e Warnings are not critical but provide information if data is missing or ac tivities are not possible Warnings are preceded by the icon iis e Exceptions are critical and indicated by the icon 4 Graphical Section In the graphical section you can design the graph As mentioned above you can drag and drop important or new cost factors to the graph section Influences between the factors can be drawn by selecting the Influence el ement in section 1 Functions on the tool bar or and menu bar L New creates a new graphical environment in a new tab It is possible to have up to 10 different tabs graphs at one time Open loads a graph in a new tab and its graphical environment In con trast to the New function the Open function loads a graph from a file which contains factors and influences Le Check By pressing this button the system verifies the designed graph and informs you in sections 3 and 4 whether the graph is consistent with syntactic modeling rules In case of inconsistency the warning icon s ii appear s For more information about rules see the chapter Validation below Accept The accept button is enabled only if you pressed the check but ton before and the graph does not
12. can be stored in an XML file at any point in time This is especially useful after the validation of the model is completed successfully To export the current COBRA project model press the Export CoBRA Project button The following dialog appears Save in S projects My Recent Documents gt File name My Network Places Files of type CoBRA Project XML cobra xmi In this dialog you can export our model as a CoBRA Project XML file cobra xml Remark Modifications made during the different steps of model develop ment are stored directly in the current model Modifications also remain if you leave the actual COBRA project i e you do not need to store the changes in the model explicitly However if you want to store a certain state of the CoBRA project model or exchange a project model with other per sons you can export the current state of the project model as an XML file You can also use the export function to migrate our models to a newer Co BRIX version Copyright Fraunhofer IESE 2011 Usage of COBRIX Import Cost Factors B 6 Import Cost Factors By pressing the Import Cost Factor button you get to the following dialog ffnen PER Suchen in la costfactorfiles v ems O aa CostFactors Mumbers chf Zuletzt verwendete empty cFF Dokumente Desktop Eigene Dateien Arbeitsplatz Dateiname CostFactors Example cht Netzwerk u
13. dialog shows all experts and whether the experts already have estimat ed the influence of the cost factor variables as multipliers Copyright Fraunhofer IESE 2011 22 Usage of COBRIX Collect Multiplier Data The multiplier estimations of each expert can be modified and removed To collect the multiplier estimations for an expert or to modify them the follow ing dialog is used Variable Mame PERS 2 1 Applicati PERS 2 2 Platform PERS 5 Project Ma PROC 1 Disciplined FROC Z Customer PROD 2 1 Difficult PRODI Important PROD 3 Implement PROD 4 Software PRODZ 2 Require PROJ 1 Clarity of PROJ 2 Developm Set Modify Multiplier The dialog shows all cost factor variables that have to be estimated by the selected expert For each variable the low most likely and high value that were provided by the selected expert are presented You can modify the values by double clicking on the corresponding line amp Add Modify Multiplier PER Variable Name PERS S Project Manager s Experience and skills Question Project Manager s Expenence and skills L Answer Extreme Inferior Hultiplier Data in full for this Variable Low o Most Likely 930s High 100 23 Usage of COBRIX Collect Multiplier Data The dialog shows the name of the cost factor variable a
14. possibilities for the stated question According to the selected scale type the predefined an swer for an extreme case is shown e The Add and Change buttons can be used to define new scale types or modify existing ones More details are provided in the next subsection B 12 5 Add Modify Scale Types If you need a scale type that is not available yet you can add it by pressing Add or Change at the bottom of the dialog The dialog that appears allows choosing to add a new scale type or modifying removing an existing one 19 Usage of COBRIX Build Causal Model amp Scale Type Management DE Name Agreement Evaluation For a scale type its name and the four answer possibilities from nominal case to extreme case have to be provided Modify Scale Type BE Hame Frequency Answer Possibility 0 Nominali Mast of the T ime Answer Possibility 1 Occasionally Answer Possibility 2 Infreguently Answer Possibility 3 Extreme Rarely C Cra B 12 6 Modify Causal Relationships A dialog containing the properties of an Influence appears when you right click the influence or double click it with the left mouse button Copyright Fraunhofer IESE 2011 20 Usage of COBRIX Build Causal Model Positive Ci Negative Why does the cost factor have this influence e In general
15. By pressing the Synthesize Multiplier button you get to the following dialog Synthesizing Multiplier Data DE Choose Variable whoes Multipliers should be sythesized Variable Mame Total Count Done PERS 2 1 Application domain e PERS 2 2 Platform experience PERS S Project Manager s Exp Lo R aa u requireme SD PROL 2ZLUS sto omer partic ipatio m FROD 21 Difficulty of require BE PROD 1 Importance of softwa PROD 3 Implementation of re PROD 4 Software performant PRODZ 2 Requirements stability ie PROJ 1 Clarity of project tea B E PROJ 2 Development schedul asic 2 Synth BRE Done In this step you can decide which experts estimations should be used to simulate the influence of each cost factor This allows us to remove individ ual outliners before the calculation For this purpose the first dialog shows a list with all cost factor variables and allows choosing a variable Pressing the Synthesize or Modify button opens the following dialog Multiplier Data estimated for PROC Customer participation EX Deselect Experts whose Multipliers for this Yariable should not be used Expert Experience M Distance Used Peter M ller Michael Urn G nther Gelb Sepp Ernst Alex Trost Paul Panzer Willi Klein Werner Breit Karin Stark Lars Ritter R diger Zimmer J rgen Schlegel rn ae
16. For each past project you have to enter the project name project size and project cost In addition you have to answer the questions obtained from the causal model quantification to determine the difficulty of the project Add new Project Elicitation PER Ti Project Name Project Fx18 Project Size kLol 300 Project Cost PH 4200 PROD 2 1 Difficulty of requirements defintion It is difficultto define the requirements 2 Strongly Disagree Disagree i Agree 2 Strongly Agree PROD 2 2 Stability of Requirements Rarely Infrequent Occasionally Most of the Time PROJ Development schedule constraints In the project exists a disciplined requirement management Strongly Agree O Agree Disagree Stong Disagree Remark In the actual version of CoBRIX for each question you can also find a button named Details The button allows us not only to select one of the four answer possibilities but also to provide probabilities for each an swer possibility For example if you are not sure whether answer possibility 1 or 2 is correct for our project you can press the Details button and pro vide a 50 percentage probability for both answer possibilities 27 Usage of COBRIX Validation of the COBRA Model B 17 Validation of the COBRA Model 28 By pressing the Validate Model button you get to a dialog with several tabs Regression is optional
17. actors for example in later COBRA projects B 9 Expert Management By pressing the Add Modify Expert button you get to the following dialog Expert Management Expenence Hprojects vee SSS SS SSS Usage of COBRIX Cost Factor Ranking by Experts This dialog allows managing the participating experts in our COBRA project You can add new experts with their experience level to the COBRA project you can modify existing experts and you can remove them from the CoBRA project Remark You can only remove an expert from the project if the expert does not participate in any COBRA estimation activity like the ranking of cost fac tors or multiplier estimation B 10 Cost Factor Ranking by Experts 10 By pressing the Ranking by Experts button you get to the following dialog Ranking of Oranisation Relevant Cost Factors by DE Expenence Hproje Ranking done Reinhart Werner Remark If you have no expert in our project you find an empty table You can add required experts in the Expert Management dialog In this process step the experts can rank the organization relevant cost fac tors If an expert has completed his ranking this is automatically marked by a check G nther The rankings can be modified or removed For experts who have not ranked the cost factors you can start the ranking process by pressing the Rank button Copyright Fraunhofer IESE
18. amount of attention given to ensure that software products are easy to understand and modify Could be achieved through actions such as information hiding modularity in design completeness and traceability of life cycle documentation and the recording of design rationale Extreme Case Importance of software maintainability Done This dialog shows all cost factors that are mentioned as relevant in the scope of the CoBRA project and should therefore be ranked by the experts When selecting a cost factor its definition and its extreme case worst case are shown You can remove cost factors that are not relevant in our context In addition existing cost factors can be modified and new cost factors can be created Copyright Fraunhofer IESE 2011 Usage of COBRIX Export Cost Factors B 8 Export Cost Factors By pressing the Export Cost Factors button you get to the following dialog Speichern Speichern in la costfactorfiles v pm i 2 CostFactors Example cFF CostFactors Mumbers chf Zuletzt verwendete empty cFF Dokumente Desktop Eigene Dateien Arbeitsplatz Dateiname new CostF actor List Netzwerk umgebn Dateityp CostFactorfile l cff w Abbrechen In this dialog you can export our modified list of cost factors as a cost factor file cff in XML format You can reuse this modified list of cost f
19. and shown only if a regression model is used to de termine productivity see productivity models in configuration file section for details The fist tab Cross Validation shows the results of a cross validation per formed with the past project data i e each past project is estimated based on an estimation model built with the remaining past projects The results of cross validation are presented in a table and aggregated in a box plot of their relative estimation errors In the table column Estimated the estimates for the project cost can be found the column Actual presents the actual project costs the column Residual contains the difference between estimat ed and actual project costs and the column Rel Error shows the resulting relative error RE of the estimate The box plot is based on these relative error values In this tab you can also decide whether the model is accurate enough for productive use i e only if you checkmark The model was validated and is con sidered as accurate enough the model can be used to simulate new projects for the purpose of cost estimation risk analysis and benchmarking On the other hand as long as the checkmark is set you cannot modify the model To modify the model you must remove the checkmark again Validation of CoBRA Model Estimated ActualCost Residual Rel Error 2108700 1823700 _ 285000 2253349 2105000 148349 26915 39200
20. d If the vertex is a variable then the category field must not be de fined e Rankings provide statistical data based on the results of the previous step Expert Ranking in the CoBRIX tool The information cannot be edit ed Newly created cost factors variables do not contain ranking data e Inthe graph we differentiate between cost factors variables and the dependent variable i e the project effort cost Each graph needs ex actly one dependent variable e Definition and Extreme case store detail information about the fac tor variable but they do not have an influence on the representation of the factor in the graph B 12 4 Quantify Factors or Variables in the Causal Model A dialog containing the properties of the selected Factor Variable appears when you right click the factor variable or double click it with the left mouse button The tab Quantification contains all information relevant for variables Copyright Fraunhofer IESE 2011 18 Usage of COBRIX Build Causal Model Factor Variable Properties Description Quantification Question Scale type Agreement Answer for an extreme case project Strongly Disagree Addl e The Question provided here will be answered in a subsequent step for each historical and new project in order to characterize the project with respect to the variable e Scale Type is used to provide predefined answer
21. estimations provided by a certain expert decreases or in creases model prediction accuracy 31 Usage of COBRIX Validation of the COBRA Model 32 Analysis of Expert Influence Box Plot RE Removed Expert Frediz5 Median RE TORange Re 2 amp 0 E01 Peter M ller Experience 20 0 15 0 75 0 03 0 22 02 02 e6 E11 R diger Zimmer Experience 0 stdM Current Model no Expert is removed GE ons ME 0 23 Variables with a green table background improve model prediction accuracy whereas variables with a red table background decrease the accuracy of the model prediction The line with the results of the current model serves as a reference to quantify the improvement potentials by the estimations of each expert Pressing the Go for Iteration allows jumping back to the causal model building step so you can modify and enhance the causal model and its quantification to improve prediction accuracy The modifications should con sider the results of the advanced analysis performed Leave one out Varia ble Expert Validation Copyright Fraunhofer IESE 2011 Usage of COBRIX Enter New Project Data B 18 Enter New Project Data By pressing the New Project Data button you get to the following dialog Projects for Estimation Done This dialog shows all projects that can be estimated with the help of our val idated COBRA model You can move finished projects to the past projects to extend
22. ettings The properties below allow you to access a DB Server on another computer or with another username and password This can be helpful if you want to use an existing DB Server on a dedicated server machine However we recommend using the integrated Derby DB provided This property names a database that is available on the server lt entry key DBMS UNIDB gt DERBY _ JDBC lt entry gt This property names the database driver lt entry key DBMS DRIVER gt org apache derby jdbc EmbeddedDriver lt entry gt This property locates the database server lt entry key DBMS_ URL gt jdbc derby lt entry gt This property names a database user with root rights lt entry key DBMS_ USER gt empty lt entry gt This property provides the password for the named user lt entry key DBMS PWD gt lt entry gt 41 Configuration File Debug Settings C 5 Debug Settings These properties are for developers who modify or debug CoBRIX They al low printing exceptions and SQL statements on the console If this property is set to true all exceptions are shown lt entry key PRINT EXCEPTION gt true lt entry gt If this property is set to true all SQL insets update and delete statements are shown lt entry key PRINT SQL gt true lt entry gt If this property is set to true all SQL select statements are shown lt entry key PRINT SQL SELECT gt true lt entry gt C 6 Productivity Model
23. for further analysis Size is provided in the defined size unit actual cost in the defined cost unit cost overhead in percentage and actual and nominal productivity in the cost unit size unit Validation of CoBRA Model i Regression Improvment Project Size Actual Cost Cost Overhead Actual Producti Nominal Produc CoBRA 01 2062 1450 0 279 1 451 CoBRA 02 2150 623 0 319 1 458 CoBRA 03 2185 336 0 387 1 805 CoBRA 04 2192 266 0 402 1 985 CoBRA 05 2246 455 0 273 0 986 CoBRA 06 2292 437 1 757 CoBRA 07 2298 1 246 CoBRA 08 23079 1 475 CoBRA 09 2314 1 254 CoBRA 10 2322 1 282 CoBRA 11 2359 f 1 565 CoBRA 12 2373 1 992 Done 29 Usage of COBRIX Validation of the CoBRA Model If the productivity model is built with the help of linear regression the follow ing optional tab Regression contains a diagram with the data points histor ical projects and the resulting regression line Validation of CoBRA Model Linear Regression Graph oo o N o a ala a N 2 39 m Q S 5 8 amp Inverse Productivity gt 236 254 273 292 311 329 CostOverhead 100 The last tab Improvement allows conducting advanced analysis as well as starting an iteration to improve the existing model Validation of CoBRA Model Cross Validation Productivity P Details Regress
24. g the cost overhead distribution This property allows choosing the simulation algorithm The three available algorithms are StdMonteCarlo which is a standard Monte Carlo approach LatinHypercube an improved simulation approach which however needs more runtime per iteration LatinAypercubePO a runtime optimized version of the Latin Hypercube algorithm which is faster but needs more memory and can be used only until 640000 iterations lt entry key SIMULATION METHOD gt StdMonteCarlo lt entry gt The number of project simulation iterations performed by Simulate Project lt entry key SIMULATION_ITERATIONS gt 1000000 lt entry gt The property allows using an advanced simulation mode where probabilities of different answer possibilities are used to simulate uncertainty regarding a certain project property lt entry key PROBABILITY4ANSWERS gt true lt entry gt C 8 Other Properties The first property allows viewing UTF 8 chars in the application by overriding system dependent font selection with Unicode fonts This is helpful if using for example Japanese characters lt entry key UTF 8 gt true lt entry gt The second property allows choosing the language used in the CoBRIX dia logs The possible values are System English German and Japanese with 43 Configuration File Other Properties System being the default value that automatically chooses the language of the o
25. he elements Factor Variable and Influence There are two option to add a new factor or variable to the graphical section 4 One can either select the box Factor Variable in section 1 and click on any point of section 4 or one can drag and drop the box Factor Variable to section 4 As soon as at least two cost factors are in section 4 you can connect nem with an influence or an edge For this you need to select the Edge or Influence button in section 1 and move the mouse cursor to the center of the cost factor so that it is surrounded by green border lines Press the left mouse button and pull the line to the connecting factor As soon as the cost factor is also surrounded by a green border you can release the left mouse button Now the two factors are connected by an influence relation ship 2 Cost Factor Section This section presents the cost factors that were defined and ranked in the previous stages of the CoBRIX tool Each cost factor can be dragged and dropped from section 2 to section 4 At the top of section 2 in the combo box you can select a specific category Usage of COBRIX Build Causal Model 16 To remove a cost factor from the factor list right click on the cost factor and select Delete Some cost factors cannot be removed be cause they have been defined as important in the previous stages of the CoBRIX tool In those cases you will get a warning in section 3 3
26. hoose Experts whose Ranking are used to calcu EX Mame Expenence Hproje Selected There you can mark the experts whose cost factor rankings should be used to create the statistics Ranking is used click on check to remove mark G nther By pressing the Ok button the statistics for the cost factors are created Copyright Fraunhofer IESE 2011 12 Usage of COBRIX Choose Most Important Cost Factors Personal Process Product Project CostFac Mean cel En Min PROD 1 Im a2 0219 moame 2 oat 028 a 3 PROD 3So 26 Dei 0558 E PAOD 4 5 0 42 O16 1 173 1 T E EEE E Kenndall s Coefficient of Concordance tes to provide evidence at a certain level of significance that the rankings are not independent Wy 0 84 is significant at 01 For each category cost factors are sorted by their importance with the most important one on top Ranking statistics are presented They contain the mean the variance the standard error the minimum the maximum and the range of the rankings provided by the previous selected experts In addition the coefficient of concordance Kendall is computed which al lows checking whether the statistics are meaningful Significance at 01 and 05 is shown W 0 84 is significant at 01 It is also shown if there is no significance Wi 0 65 is HOT significant You can select the most impo
27. ion Improvment Go for Interation Leave one out Variable validation Leave one out Expert Validation Copyright Fraunhofer IESE 2011 30 Usage of COBRIX Validation of the CoBRA Model Pressing the Leave one out Variable Validation allows comparing the influ ence of each single variable on the estimation results The results of remov ing a certain variable from the model are presented by a box plot as well as in figures The figures show whether the removal of a certain variable de creases or increases model prediction accuracy Analysis of Variable Influence Box Plot RE Removed Variable O_PROD 1 Novelty of requirements vi_PROD 2 Importance of product maintenance ve PROD S Novelty of technology v4 PROC Customer s up front input amp motivation std M Current Model no Yariable is removed Variables with a green table background improve model prediction accuracy whereas variables with a red table background decrease the accuracy of the model prediction The row with the results of the current model serves as a reference to quantify the improvement potential by each variable Pressing the Leave one out Expert Validation allows comparing the influence of the estimations of each single expert on model prediction accuracy The results of removing all estimations of a certain expert from the model are presented by a box plot as well as in figures The figures show whether the removal of all
28. ity activities E Redo restores the last undone activity activities Select All selects all factors and influences of the current tab and its graph ical environment Select None deselects all factors and influences s Zoom In zooms into the graphical environment by 120 Zoom Out zooms out of the graphical environment by 120 H Fit to Size resets the page to its visible and graphical page size 17 Usage of COBRIX Build Causal Model B 12 3 Modify Factors or Variables in the Causal Model A dialog containing the properties of a factor variable appears when you se lect and right click the factor variable or double click the left mouse button on a cost factor Factor ariable Properties Description Quantification Definition FERS 1 Consistency of project objectives The extent to which customer 2 and the project team are consistent in their understanding of project objectives le no conflicts in their interpretation of Category the objectives Rankings Mean 1 00 Std Err Marl Min 1 00 Extreme case Marx 1 00 Range 0 00 Add Factor to Factor list Consistency of project objectives Dependent variable effort The Description tab contains all information relevant for a cost factor e The Name of the factor variable must be unique in the graph e A cost factor belongs to one Category which is defined in the category fiel
29. mgebi Dateityp CostFactorfile Lct a E Abbrechen In this process step you can import a proposal list of predefined cost fac tors If you want to start without using such a list you can load empty cff which contains no cost factors Usage of COBRIX Synthesize Organization Relevant Cost Factors B 7 Synthesize Organization Relevant Cost Factors By pressing the Synthesize org rel CF button you get to the following dialog Synthesize Organisation Relevant Cost Factors DE Name Category PERS 1 Consistency of project objectives Personal PERS 2 Level of experience and knowledge Personal PERS 3 Communication and team work skills Personal PERS 4 Project team morale ersanal FERS 5 Project Manager s Experience and Personal PROD 1 Importance of software reliability roduc PROD 2 Requirements Definition amp Stability Product FROD 3 Software product complexity roduc PROD 4 Software performance constraints Product PROD 5 Importance af software mairtainab Product PROC Disciplined requirement managem Process PROC 2 Customer participation oces PROC 3 Implementation of reviews and ins Process PROC 4 Disciplined quality management moce FROC S Quality of testing OCES PRO Clarity of project team roles and res Project PROJ Development schedule constraints Project au a a DEE nyn Definition The
30. nd the associated question with the answer for the extreme case worst case The experts have to estimate the cost overhead produced through the extreme case of the variable e g inferior skills and experience of the project manager The estimation is provided with a low value lowest possible influence on project cost a most likely value the cost overhead occurs most likely and a high value the highest possible influence on project cost The estimates are given in percent of cost overhead Example If an expert expects that the inferior skills and experience of the project manager lead to a minimum 0 increase in project costs most likely leads to a 30 increase in project costs and maximal to additional project costs of 100 he provides KHultipher Data in full for this Variable Low T Most Likely 30 High 100 Moreover if an influencing cost factor exists its extreme case is presented and the expert must estimate the influenced cost factor variable for both cases e the influencing cost factor is nominal optimal and e the influencing cost factor is extreme worst case Extreme Case of Influencing Cost Factor Disciplined requirement management Multipher Data in full for this Variable iF the Influencing Cost Factor is Nominal Extreme Low 5 30 Most Likely 20 100 High ot UU Copyright Fraunhofer IESE 2011 24 Usage of COBRIX Synthesize Multiplier Data B 14 Synthesize Multiplier Data
31. o select one of the four answer possibilities but also to provide probabilities for each an swer possibility For example if you are not sure whether answer possibility 1 or 2 is correct for the actual project you can press the Details button and provide a 50 percentage probability for both answer possibilities Copyright Fraunhofer IESE 2011 34 Usage of COBRIX Project Simulation B 19 Project Simulation By pressing the Simulation button you get to the following dialog Choose Project for Simulation DE Project Name You can choose the project that you want to simulate The simulation cre ates a cost overhead distribution based on the COBRA model and the infor mation provided for the selected project The distribution is required to esti mate project cost analyze cost risk and benchmark the project If you start simulation the following dialog which shows the progress appears Fortschritt Running Project Simulation Completed 280 000 out of 640 000 35 Usage of COBRIX Estimate Project Cost B 20 Estimate Project Cost By pressing the Estimate Cost button you get to the following dialog Estimate Project Cost PB x Project Name Project IESE G1 Probability Distribution hi Lae un co Oo j gt iz E a Lo 5 je 1966 2537 3108 36579 4251 4822 5393 5964 6535 7107 7678 Cost in PH MEAN of the estimated Projec
32. peration system as the language for the CoBRIX dialogs lt entry key LANGUAGE gt System lt entry gt The third property defines the maximum number of CoBRA projects that can be stored in the database A high number of projects can reduce application start performance lt entry key MAX_PROJECTS gt 20 lt entry gt Copyright Fraunhofer IESE 2011 44
33. rience metrit Unit Hprojecte Definition umber projects in this context First you have to enter the project name Then you can enter the context scope of the CoBRA project In addition you should decide on the size cost and experience metric you want to use The Ok button brings us to the Process Overview Dialog Copyright Fraunhofer IESE 2011 B 4 Usage of COBRIX CoBRA Process Overview CoBRA Process Overview Z ra u r Initialization pe Fraunhofe E S E Import Cost Factors pe v synthesize arg rl CF 9 0 gt Export Cost Factors y Ranking by Experts 5 5 Choose Most Important View Model m Emmen 5 v r Synthesize Multiplier Elicitate Past Projects ae ee eee Validate Model New Projects Simulation Estimate Cost Analyze Cost Risk Benchmark Projects The process overview dialog guides us through the CoBRA process All pro cess steps you can do next are shown as enabled buttons Import Cost Factors Process steps that cannot be executed yet are disabled 7 Senthesize Ord rel CF Usage of COBRIX Export CoBRA Project If you want do undo a previous process step you can use the undo button The Done button closes COBRIX The current project state is safe in the database and you can continue working by choosing the project in the start ing dialog COBRA Project Selection B 5 Export CoBRA Project The CoBRA project i e the model developed
34. rtant cost factors that you want to use in our CoBRA model PROD2Re 4 w oa o a y 13 Usage of COBRIX Build Causal Model To navigate through the individual categories Personal Process Product Project the Back and Next buttons can be used When the most imported cost factors have been marked for all categories you can store our selection and close the dialog with the Finish button B 12 Build Causal Model By pressing the Build Causal Model button you start the CoBRIX Editor It allows us to create a graph representing the causal model you use for the cost estimation The vertexes of a graph are cost factors or variables and the edges are so called influence relationships which model the dependen cies between factors variables B 12 1 Create New or Open Existing Causal Model When starting the CoBRIX Editor you have the option of either creating a new graph or loading an existing graph It is possible to open or create up to 10 different graphs Each of them is stored in its individual tab The corresponding icons are located on the tool bar B 12 2 Causal Model Editing After creating or opening a graph you can see the following sections Copyright Fraunhofer IESE 2011 14 Usage of COBRIX Build Causal Model DsBdl Zub x Sc 332 Ge E Hea arahi kiai iiss besiis fo Tame ne 1 Tool Box Section In this section you have t
35. t Cost PH The dialog shows the probability density distribution of the project s costs in the given project cost unit The mean i e average project cost is present ed as well MEAN of the estimated Project Cost PH 4716 Copyright Fraunhofer IESE 2011 Usage of COBRIX Analyze Project Cost Risk B 21 Analyze Project Cost Risk By pressing the Analyze Cost Risk button you get to the following dialog Analyze of Cost Risk PER Project Name Project IESE G1 Cummulative Probability Distribution Win budget B125 W125 amp Lis zx o a g E E sen E Costin PH Budget in PH 4000 Risk of been over Budget 56 Risk of been more than 25 over Budget 4 You can calculate the risk of being over budget or even of being 25 over budget for every given budget by entering the budget and pressing the but ton on the right Budget in PH You can also calculate how high the budget has to be so that it is not ex ceeded by a given probability To do this you enter the probability and press the button on the right Risk of been over Budget Usage of COBRIX Benchmark Projects The diagram in the dialog gives a graphical interpretation of the results Cummulative Probability Distribution Win budget E 125 5 2 g 5 a E E D AA 5393 p54 6535 Costin PH B 22 Benchmark Projects By pressing the Benchmark Projects button
36. the base for future estimations and hopefully improve the model by pressing the Move to Past Projects button You can also add new projects to the list modify existing ones or remove them To add a new project you have to press the Add button ETE PIES PER ls Project in Scope of this CoBRA Model automotive industry embedded C maintenance Before you enter data of a new project you are asked to check whether the project is or will be conducted in the previously defined context Scope of the CoBRA project After pressing Ok you can enter the name of the project and its estimated size Additionally you have to answer the questions derived from the quanti fication of the causal model 33 Usage of COBRIX Enter New Project Data Add new Project to Estimate Project Name Project wid Project Size kLoC 150 0 Project Cost FH PROJ Development schedule constraints In the project exists a disciplined requirement management O Strongly Agree Agee Ne O Disagree Strong Disagree PROD 2 1 Difficulty of requirements defintion Itis difieultto define the requirements Strongly Disagree Disagree gt Agree S Strongly Agree FROD 2 2 Stability of Requirements O Rarely O Infrequentl Occasionally Mast of the Time Remark In the actual version of CoBRIX for each question you can also find a button named Details This button allows us not only t
37. violate any modeling rules This means Copyright Fraunhofer IESE 2011 Usage of COBRIX Build Causal Model that the graph is a valid COBRA causal model that can be stored in the da tabase By pressing the Accept button the necessary data regarding cost factors variables and influences will be stored in the database and the CoBRIX Editor will be closed In the CoBRIX tool the step Build Causal Model will be marked as completed Save As soon as the graph has been modified you can save the cur rent graph to a file If the graph was not saved before the function Save As will be used automatically I Save As is similar to the Save function but allows you to provide a file name for the saved graph Ly Page Setup allows choosing the page size and orientation as well as the printing settings Print sends a print job to print the current graph Exit closes the CoBRIX Editor Modifications will not be stored in the database amp Cut cuts the selected cost factors or influences from the graph and stores them on the clipboard Ti Copy copies the selected cost factors or influences and stores them on the clipboard 5 Paste inserts all cost factors and influences stored on the clipboard into the currently selected tab and its graphical environment X Delete deletes the selected cost factors and influences Undo undoes the last activ
Download Pdf Manuals
Related Search