Tips on using this forum..

(1) Explain your problem, don't simply post "This isn't working". What were you doing when you faced the problem? What have you tried to resolve - did you look for a solution using "Search" ? Has it happened just once or several times?

(2) It's also good to get feedback when a solution is found, return to the original post to explain how it was resolved so that more people can also use the results.

Monte Carlo Risk Analysis

41 replies [Last post]
Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Sometimes. when performing Monte Carlo risk analysis I get a message ''no paramters for modeling". Can anyone tell me the exact steps for performing the Monte Carlo risk analysis.

Thank you,

Best Regars,

Bogdan

Replies

With formulas that set different numbers in user field that is made signal you may create different colors for different activity types but no patterns. Colors on the phase may reflect user defined colors of its sub-phases. But it makes sense to discuss this in the separate topic, it is not connected with MC analysis.
With formulas that set different numbers in user field that is made signal you may create different colors for different activity types but no patterns. Colors on the phase may reflect user defined colors of its sub-phases. But it makes sense to discuss this in the separate topic, it is not connected with MC analysis.
Rafael Davila
User is online Online
Joined: 1 Mar 2004
Posts: 4723

You are right, even when for signals alone it does not makes much sense it does when used as to dynamically change bar colors, the tweaking I rarely use.

I miss being able to define colors/patterns by activity type, the most relevant for me would still be the Hammock.

Maybe with formulas and signals hammock color can be tweaked, but it is too complicated to even search for it.

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163
Perhaps partterns could also be added to signals. This way the users could have even more flexibility.
Rafael Davila
User is online Online
Joined: 1 Mar 2004
Posts: 4723

Bogan,

I am not much into tweaking colors other than manually. If in addition to colors there were patterns I would get more interested as photocopies and print out from my B&W laser printer would make sense.

When I was a SureTrak user my favorite tweaking, and perhaps the only was to setup hammock bars to be light grey with diagonal lines. Something simple and always the same, more on the practical side.

I would like the following option box to provide for tweaking color of activities by Type DPH and to provide pattern options in addition to color and would avoid use the other complicated methods, only manual color for a few activities I want to highlight. 

 

 photo 1-11-20147-51-33PM_zpsa17f6437.jpg

Best regards,

Rafael

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Rafael,

I saw in one of the Spider Team publications ("Tools and Techniques of Project and Portfolio Management with Resource Constraints") project phases using two colors, I was wondering if you know if this is possible in the program and how. I belive the "Use subphases colors to draw phases" option under "Gantt Options" may be the way to do it, but I haven't figured how to use it. Any suggestions?

Regards,

Bogdan

Rafael Davila
User is online Online
Joined: 1 Mar 2004
Posts: 4723

Bodgan, 

I have an SSD Drive and the computer reads the file from the SSD, this loading of file from SSD to RAM is where I suppose most of time saving is as I believe Spider runs on RAM similar to Pertmaster, if there is still some swapping then extra time savings might be experienced, but this is outside my area of expertise.

The best way to compare would be using 2 similar computers one with HD and another with an SSD against similar MC file and same number of iterations.

Part of the enhanced experience might be with the automatic backup that happens on the background, with an SSD it should be much faster.  I would not be surprised if during MC run the backup utility is on hold. 

Suggest using previous version support for your resource leveling algorithm rather than optimization algorithms. For project stability I use optimization algorithm at the beginning and when in trouble. In my jobs what makes the difference are the functionalities in Spider not available in other software such as variable quantities and workloads, but this should also add overhead to the calculations. 

Perhaps Vladimir can give us some light as you might be correct in suspecting after file loading there are no substantial gains in computing speed between HD and SSD as most will be performed on RAM. In such case RAM speed, amount and cache might make the difference. 

My best wishes for the starting year.

Rafael

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir, Rafael

first of all I would like to wish you a very productive and good year.

Second, I have a few questions for Rafael regarding his SSD. I remeber you writing that with a SSD the MC is very fast, which makes a lot of scense. What I don't know is if you read the file from the SSD or from a mechanical drive. Since the information is stored into the file only after saving the project logic dictates that it doesn't matter if the file is on the mechanical drive as long as the program runs from the SSD.

Also I was wondering about speed increase in other calculations such as resource leveling, are there any improvements?

Best Regards,

Bogdan

Don't hesitate to ask any questions.

Regards,

Vladimir

envision global
User offline. Last seen 5 years 19 weeks ago. Offline
Joined: 11 Dec 2013
Posts: 10
Groups: None

Generally i search for such information it is very useful and helpful for young people like me to struggling to make a carrier

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

That clears the question about triggers having probility equal to zero or empty cell.

Regards,

Bogdan

Bogdan,

Triggers represent risk events, and there could be secondary risks – so other triggers may follow initial triggers as may be seen in Trigger Chart.

Triggers have probability to happen and if happened may initiate different network branches with different probabilities.

Trigger state is selected manually but this selection does not change Monte Carlo distribution that deals with probabilities of different states (and branches).

In deterministic networks triggers behave as switches with selected states.

Empty trigger probability is the same as zero.

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Greetings Vladimir,

I took a closer look at Trigger Activities in order the understand how they work and influence MC analisys. I used a simple project with only one Trigger Activity, with variable probabilities and 3 chains from the Trigger Activity. My next observations reflect the behaviour of the Trigger Activity and it's chains regarding the MC Probability Distribution, more precise it's maximum and minimum values.

Initial data:

Duration of Project for each of the 3 estimates:Optimist Estimations: 18 DaysProbable Estimations: 30 DaysPessimistEstimations: 48 Days 

Observation 1: If trigger has probability less than 100% then there must be a link to the another activity that is outside the Trigger condition. Also, if multiple triggers that are in a sequence (one leads to another) then the triggers can be linked like a chain as long as the last trigger with probability < 100% is linked to a trigger with 100% probability or another activity that is not a trigger. This check is performed when running MC analysis.

 

Senarios:

Scenario 1: Inactive Trigger (Probability = ''; State = '')

Minimum Value of the Distribution: PbMinD = 27.58 Days

Maximum Value of the Distribution: PbMaxD = 34.81 Days

Observation 2: In this scenario the trigger behaves as if the trigger is a simple milestone and there are no trigger conditions.

Scenario 2: Trigger Probability = "" or 0.00

Trigger State: Chain 1

Behaviour, Trigger = Milestone/Switch Activity

Duration of Chain 1: DurChain1 = 10 Days

Minimum Value of the Distribution: aproximately(PbMinD+DurChain1) = 37.45  Days

Maximum Value of the Distribution: aproximately(PbMaxD+DurChain1) = 44.10 Days

Trigger State: Chain 2

Behaviour, Trigger = Milestone/Switch Activity

Duration of Chain 2: DurChain2 = 20 Days

Minimum Value of the Distribution: aproximately(PbMinD+DurChain2) = 47.45 Days

Maximum Value of the Distribution: aproximately(PbMaxD+DurChain2) = 54.10 Days

Trigger State: Chain 3

Behaviour, Trigger = Milestone/Switch Activity

Duration of Chain 1: DurChain3 = 15 Days

Minimum Value of the Distribution: aproximately(PbMinD+DurChain3) = 43.08 Days

Maximum Value of the Distribution: aproximately(PbMaxD+DurChain3) = 49.94 Days

Observation 3: If trigger probability is empty or set to zero (0) and a chain is selected then the trigger activity behaves like a Switch activity with multiple states (Chains).

Scenario 2: Trigger Probability = 100

Trigger State = Chain 1 or Chain 2 or Chain 3 or ''

Minimum Value of the Distribution: aproximately 38.29 +/- (0.1-1 Day)

Maximum Value of the Distribution: aproximately 54.95 +/- (0.1-1 Day)

Observation 4: No observations

Scenario 3: Trigger Probability = 0.0001

Trigger State = Chain 1 or Chain 2 or Chain 3 or ''

Minimum Value of the Distribution: aproximately 28 +/- (0.1-1 Day)

Maximum Value of the Distribution: aproximately 34.82 +/- (0.1-1 Day)

Observation 5: Trigger behaves as ispected since it's probability is almost 0. It behaves similar to Scenario 1, only you can select a chain and get aproximately the same results as the ones in Scenario 1.

 

General Observation: Renaming Trigger State doesn't update in the Activity Gantt, rescheduling with renamed State, without reselecting the renamed state or any state, behaves as no Trigger State is selected.

 

Questions:

0. Did I create the Trigger and it's chains correct, or did I miss something?

1. Wouldn't it be helpful/practical to check if Trigger Probability is empty and give a warning message? Similar to the message given if Trigger probability < 100 and there is no link to another activity - Observation 1?

2. Is the behaviour for Probability = 0 and State != '' or State = '' intended? If so wouldn't it be better to set this behaviour for a negative probability?

 

Here is the file: http://speedy.sh/Vzkkw/Trggs.001.sprj

 

Best Regards,

Bogdan

 
Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Thank you,

I got stuck in my head the ideea of scenarios from the 3 scenario method, this is the reason the first example wouldn't take into account the internal financinc cost component, I deleted the coresponding activities in the other scenarios thus resulting in a greater value in the expected estimate. When I imported the data from the 3 scenarios it had no activity to assign the value to. Everything makes sense now, now both examples perform as I wanted them to.

My apologies for speaking without doing my homework. I have no experiene using MC analysis.

Thank you,

Best Regards,

Bogdan

Bogdan,

there are no scenarios in Monte Carlo.

Just three estimates of initial data, triggers with probabilities to happen with outgoing probabilistic dependencies.

Project scenarios are created for three scenarios method only.

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir,

I changed the formulas in each scenario so the calculate cost components will have different values. But since you mentioned that the MC doesn't allow formula cost components, this means that there aren't any bugs. 

This however raises another question, why aren't formula calculated cost componets allowed in MC analysis.

Best Regards,

Bogdan

Bogdan,

fixed cost component in your project is the same for optimistic, most likely and pessimistic estimates.

If other cost components are calculated by formulas they will also will be the same.

When calculate by formula is unticked your manual estimates for these components (that are different) will be used in MC calculations.

Spider does not permit to use optimistic, most likely and pessimistic formulas in Monte Carlo simulation.

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

 

Vladimir,

Regarding the second file (the one you managed to download):

1. The internal financing isn't exactly 0, it's value is 1e-10 (I gave it this value since I though that the MC wouldn't work if it was 0). Also if a cost components is 0 why wouldn't it be considered in the MC analysis?

2. I tried changing the internal financing component so it would have a greater value, like 100, still nothing.

Regarding the first file (the one I uploaded a second time, containing a archive called 'Monte Carlo'):

1. I created 2 scenarios in order to see the differences.

2. The issue is in the scenario called opt2.

3. The problem is that if a cost component remains fixed throughout the scenarios and other cost components that are computed by formula vary (I changed the formula in each scenario), the MC will not show any of the cost components/cost centers containing the cost centers(this is in the opt2, cmp2, pes2 files).

If I don't change the cost component called 'fixed' but I untick the 'Calculate by formula' then the 'formula' calculated cost components will show in the distribution menu. The formulas seem to be in some way emebed in such a way that if I untick the 'calculate by formula' box of the 'Formula 1' cost comp then the next 2 (Formula 2,3) will show even though their 'calculate by formula' boxes are ticked(this is also in the opt, cmp, pes files)

On the other hand if I vary the cost component called 'fixed' by even a very small value such as 1e-10 then the formula calculated cost components will be available(this is in the opt, cmp, pes files).

 

If any of this doesn't make sense I could record a small movie of the steps I performed in order to better explain, maybe I'm messing up somewhere.

Best Regards,

Bogdan

 

Bogdan,

in your project Internal financing is zero.

I also did not find problems in other (Opt) files. Cost component distributions are shown.

Looks like I miss something, will look again tomorrow.

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Here is the first file: http://speedy.sh/ZzpTc/Monte-Carlo.rar

Also, regarding the second file, the one you have downloaded, I tried selecting all the cost components and cost centers for MC analysis and when trying to view the probability distributions they are available for all cost centers and cost components except Internal Financing (even tried to run the 3 scenario analysis first and importing the data - though I don't belive this should have any importance).

I don't know why the analysis ignores this cost component, also the cost centers have this cost component, how does this affect the results for the cost centers?

 

Best Regards,

Bogdan

Bogdan,

I tried the last opt file. 

In this file Cost and Internal Financing were selected forrisk analysis but other components Payments and Revenue were not selected. When I selected them too all reports on cost distribution were available.

Unfortunately I did not download your previous file and now it is not available. Can you upload it once more?

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir,

I was wondering if you had had the chance to take a look at the examples I posted a few weeks ago about some possible MC bugs? I have downloaded the new version but when I ran the MC analysis on one of the examples I got the same results, to be more precise I didn't get any results.

Thank you,

Best Regards,

Bogdan

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Thank you,

also when you have the time please take a look at the example I uploaded in the previous post since I belive there may be a bug regarding that perticular topic.

Best Regards,

Bogdan

Bogdan,

thank you for your feedback.

I am very busy these days and will reply at the end of April. I am sure that you found some bug, we will look and fix it soon.

Best Regards,

Vladimir

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir,

I belive I have found another mishap of the MC analysis. I don't know how to explain this one, the only thing I can say for certain is that the MC analysis simply ignores all cost components except one. I don't know why?!

I have uploaded the file here: http://www.speedyshare.com/QQPEZ/Internal-Finance-Risk-Analysis.rar

Best Regards,

Bogdan

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

 

I realized the last 2 post are pretty much nonsense so I tried to narrow down my observations/questions. Hope I was able to formulate them clear and concise this time.

Observation 1: Consider a cost component doesn't vary in the 3 senarios and other cost components, which are computed using formulas and are based upon the non-variable cost component. If in the scenario (the actual physical spider file, not the option in the MC analysis window) the calculate by formula option is ticked then the MC analysis will run but when I want to see the distribution it looks like non were calculated. If I untick the calculate by formula then everything is OK. Also if the non-variable cost component varies in each senarios by as much as lets say 10e-9 then everythig works perfectly and there is no need to untick the calculate by formula option. Also when I tried comparing the distributions of the 2 projects (variable and unticked, non-variable and ticked) I could do it only for the 'duration parameter'.

Observation 2: If a cost component/cost center is 0 in one senario (optimistic) the MC will not run for that cc/CC (if the mentioned cc/CC is the only parameter in the analysis then I recieve the message 'no parameters selected' - this I belive is intended as a failsafe.

My questions are: Are my observations correct or am I missing somethig? If this is all normal, Vladimir, could you explain the logic behind it so I can better understant how to run MC using Spider?

I have attached 2 versions in the MC analysis - the files in which I performed the MC analysis is the 'MC-opt' and 'MC-opt2' files. In the 'MC-opt' file the values of the fixed cc vary from one scenario to another and in the 'MC-opt2' they don't vary. And one for the second observation - Test2.

 

Files: http://www.speedyshare.com/tNxuD/Monte-Carlo.rar

 

Thank you,

Best Regards,

Bogdan

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

I did a little more experimenting with the MC, it think I got the hang of it, but the part about the formula computed cost components remains.

Even though it's not a big deal to switch the formula off when doing the risk analisys and check it again for planning and cost computation I was wondering if the behaviour was correct or I am still doing something wrong.

Best regards,

Bogdan

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Greetings,

today I tried using MC analysis for a schedule loaded only with costs (costs per unit volume). In the three scenarios I changed only the formulas of some cost components and recalculated the project, but when I ran the MC analysis (only on those cost components computed from formulas and cost centers which contain those cost components)  I could not select any distributions for cost comp or cost center like they were never performed, even though the analysis was performed.

I tried creating a small project with 2 cost components and one cost center, one being computed using a formula and what I have noticed is that if in the scenario in which I perform the MC analysis the formula computed checkbox is ticked then I am not able to view the distributions of either cost comp or cost centers, even though the analysis is performed.

I haven't had very much time to play with the MC analysis so if I'm doing something wrong please tell me.

When I perform the MC analysis I do the following:

0. Check the perfrom MC for all activities and phases;

1. Perform the 3 scenarios risk analysis;

2. Import MC data from 3 scenarios;

3. Run MC analysis.

Thank you,

Best Regards,

Bogdan

Bogdan,

Monte Carlo scripts are not ready yet. Let our customers to get an experience of using MC before the automation of MC simulation.

As I wrote earlier MC for large projects is less practical than 3 scenarios method if to apply success probability trend analysis.

Best Regards,

Vladimir

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir,

it seems there isnt'a script to set up the Standard Fields for MC simulation, only cost comp, cost centers, materials, mat. sets/centers and user fields.

Best Regards,

Bogdan

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir,

I have downloaded the new version yesterday and saw that the issues have been fixed but I didn't get the chance to post.

Thank you,

Best Regards,

Bogdan

Bogdan,

download new version. Now this and another problem with multiresouces and skills are solved.

I am very busy these days, will participate in the discussions later.

Thank you and Best Regards,

Vladimir

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir, 

I undestand what you have written, but it doesn't help me. I was asking for which columns I can use the Monte Carlo checkbox. From what I have tried so far I could not show the Monte Carlo Probability Column for cost centers only for cost components. Here is a very simple example (1 Cost Components -> Cost, 1 Cost Center -> Cost Center 1)


Here is the link to the project files: http://speedy.sh/2YAHa/MC-TEST.rar

Rafael Davila
User is online Online
Joined: 1 Mar 2004
Posts: 4723

I suppose the Drag Cost can be defined by user defined fields and then used in Monte Carlo if there is risk distribution associated with such values. These are estimated values by the user while Drag is a computation solved using the CPM algorithms.

Stephen Devaux
User offline. Last seen 2 years 4 days ago. Offline
Joined: 23 Mar 2005
Posts: 603

Hi, Vladimir.

 

Just curious -- is drag one of the parameters you can select in Spider's Monte Carlo? How about Drag Cost and True Cost?

 

Fraternally in project management,

Steve the Bajan

Bogdan,

when you check this box Spider Project shows MC columns for selected project parameters (not only remaining durations but also optimistic, most probable and pessimistic estimates of remaining durations, etc.).

You can select and show columns with probabilities to achieve target values for materials, cost components, etc. if to select corresponding groups of data like in the picture below:

 photo 0Selectcolumn_zpsc1779f02.jpg

Probability may be shown even when you did not chose to show three MC estimates for selected parameters.

Best Regards,

Vladimir

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Vladimir,

I have been experimenting with the MC analysis, and there is something I cannot figure out. In the show columns menu on the right side of the menu, at the bottom there is a a checkbox entitled Monte Carlo. Pretty much I understand for what it is used (to show results form MC analysis), what I can't figure out is which columns are affected by checking this box, so far I could only show fields in the Standar Field Groups (Start [Probability][Monte Carlo], Finish [Probability][Monte Carlo], Total Cost [Probability][Monte Carlo]). I could not show MC for cost centers probability, cost components, materials/materia sets (didn't try yet).

Could you tell me how this works.

Best Regards,

Bogdan

Rafael Davila
User is online Online
Joined: 1 Mar 2004
Posts: 4723

 photo selected_zps7264b687.png

This message may appear if you did not select any parameter for Monte Carlo simulation in Monte Carlo Options.

Bogdan Leonte
User offline. Last seen 1 week 2 days ago. Offline
Joined: 18 Aug 2012
Posts: 163

Unfortunetly I have done this step, after this I performed a 3 scenario analysis and imported the 3 scenario data into the monte carlo analysis, because this is the only way I could make work it for any project, and I still get the previously mentioned error.

Most probably I forgot to do something, I'll keep checking.

Thank you,

Regards,

Bogdan

Rafael Davila
User is online Online
Joined: 1 Mar 2004
Posts: 4723

Make sure the Opt, Exp and Pes Remaining Duration fields on non-milestones have defined values if Perform Monte Carlo is marked as Yes, also if  under the field Perform Monte Carlo none is marked as Yes then no MC run will be executed.

 photo perform_zpse08d9cfe.png

Let me know if this solve the issue, Monte Carlo is so new I am learning by trial and error until Spider Development Team provides us with some reference. Still I am finding it easy to figure it out as it is in many ways it is similar to the 3 scenarios method.

Good Luck