Probabilistic Risk Analysis - Science topic

Hi Arman

I confirm the previous answer by Zhongze Xu / Steve. It also depends on the way the DSHA spectrum is computed, i.e., how the aleatory variability σ of the GMPE is taken into account ( 0 σ, 1 σ, ...)

Best regards

py

Inhibitors generally have a negative effect, but few people take advantage of them for a new start

Many thanks, anyway Gianluca Regina

Please get it.

You may use the approach for regression analyses. That should work.

Thank you very much for the advice. I am a pharmacist and am not very good at statistics. Therefore I used SPSS to handle. Please show me steps/guide materials to do it? Thank you :)

Hello Majid

Just a few words to answer you, although I was not part of the NGA / NGA2 projects. Essentially, from a methodological viewpoint, these NGA GMPE are not different from the previous ones you quote. They may however considered are significant improvements for several reasons :

- very careful gathering of a common set of high quality strong motion data from various parts of the world, and covering an as brand as possible range of magnitude and distances, together with the corresponding Metadata regarding the source characteristics (rupture extension, fault orientation, moment magnitude, etc.), the propagation characteristics (different source-receiver distance metrics, potentiality of source- receiver configurations), and the site characteristics (the poorest though in my opinion, with only VS30, and most often only inferred and not measured - but with a clear indication of the origin of VS30 value). Personally I do no trust that much the Z1 / Z2.5 values (depth at which the S-wave velocity exceeds 1 or 2.5 km/s).

- derivation of a parallel set of numerical simulation results (with 1D codes for horizontally layered media) in order to constrain the non-linear part of the site response

- on this common data set, several groups of authors have made their own section, and developed their own a priori models for the source, path and site effects, with for some of them very sophisticated models to (try to) account for peculiar effects (directivity, rupture mechanism, NL site response, + what is ambiguously called "basin effect" corresponding indeed to deep deposits effects, etc.)

- a very careful estimation of the aleatory variability and its systematic separation in two terms ( within and between-event), with or without additional dependence on magnitude / distance / site conditions.

These NGA were developed over several years with extensive discussions between many participants, and can thus be considered as being the best State-of-the-art at the time they were developed.

Hoping to have brought useful answers, and hoping also NGA/NGA2 authors will complement and correct me !

pyb

I believe setting confidence intervals is more of an art than an exact science. It really depends on how important it is that something remains within the confidence interval, but there are no real quantitative standards.

If going over budget almost certainly means default you may want to base risk reserves on a high percentle. If default is not imminent, like with governmental institutions, the percentile may be set lower.

If the building doesn't have an historical data of the latest fire scenarios, you should consider same kind of building statistics. FRA Analysis from SFPE could help a lot.

It depends on what you are scanning. If you are attempting to use the central limit theorem then you must verify that the underlying hypotheses are indeed true if you want to prove something. Otherwise, you have to make assumptions and if too many than the proof is not very reliable.

There are other issues as well depending on what you are trying to measure. If it has some underlying geometric structure then you may have to prove a uniqueness of approximation theorem. Otherwise, your increased sampling may converge to one of many different things and replication may be a problem. For example if you take X-rays of increased frequency of some unknown object from a single point source, increasing frequency allows you to construct some approximate shape of the unknown object but what you construct could be one of infinitely many different shapes producing the same measurements. However, if you use two somewhat independent point sources AND you know the shape is convex then it is known that increasing the number of measurements for X-rays, at some uniform angular spread from each source, will lead to approximate shapes which have to converge to the true shape of the object.

The point is, your question is a bit too general. Need more specifics.

FMEA itself is a deterministic way of classification of failures and consequences! however, if you put the amount of probability of any occurrence within the blocks in order to find the final reliability, it is considered as a probabilistic method.

as observed by Matthew:

"At 3:36 in the video is states "the only shape that has a marked effect upon the way a rock rolls is demonstrated by this elongated piece of columnar basalt. Its length gives it eccentric action." By this I think they mean it inhibits its ability to roll. If you have ever tried to roll a fireplace sized log down a natural slope, you will have observed the difficultly that is being referred to."

An elongated rock (or log) is unable to change direction as easily as the natural slope changes aspect which cause it to wobble or bounce and slow its momentum. Not sure how/if you can model that effect.

Elongated rocks also have more of a tendency to break apart as they travel down a slope.

It is fortunate that the rockfall trap is effective for each of your simulations and not dependent on this issue.

good luck

Dave

Your first statement is very insightful. Theoretical scientists like us always like to reduce the real world to a more manageable model which necessitates simplifying assumptions. So first verify your assumptions, but in the real world, as Einstein says - a mathematical model will always be uncertain! see https://www.linkedin.com/pulse/risk-assessment-21st-century-fit-purpose-david-slater/

If you are interested, you can read "Managing Price Risk for an Oil and Gas Company". Let me know.

Please let me know if the following reference and site are useful to you:

Quantifying risk:  measuring the invisible

Conference Paper Risk Management 10 October 2015

By Meyer, Werner G.

citation:  :

Meyer, W. G. (2015). Quantifying risk: measuring the invisible. Paper presented at PMI® Global Congress 2015—EMEA, London, England. Newtown Square, PA: Project Management Institute.

check the following site to see if the article is freely downloadable:

Dennis

Dennis Mazur

Not sure I have understood the question as it seems to self contradict (which may be more my lack of insight) so i will rephrase how I think you mean. 

You start with data D1 and D2 as inputs into systems S1 and S2 respectively and you obtain outputs O1 and O2 respectively.

D1 and D2 are highly correlated. O1 and O2 are not well correlated. So S1 and S2 are diverging in their behaviour.

You expect S1 and S2 to behave similarly, therefore you were expecting O1 and O2 to be highly correlated.

You want to know how to mathematically transform S1 and S2 so that the produce better agreement.

Is this correct?

If not apologies for misunderstanding.

If yes, is your D1 and D2 signal based or at least consist of ordered variables where neighbouring variables have a strong covariance?

Hi, I am not familiar with gamlss package nor beta inflated distribution but I think it should do the job. From your reference, p0 = nu /(1+nu+tau) & p1 = tau/(1+nu+tau).

In raw in the piece of code below, you will find the lines which should solve your problem.

Best regards,

Geoffrey

## Code for Janka V. ##

rm(list = ls())

require(gamlss)

set.seed(28122016)

rho = 0.2

N.obs = 2000

X1 = rnorm(N.obs,0,1)

X2 = rho *X1+sqrt(1-rho^2)*rnorm(N.obs,0,1)

Y = matrix(X1^2-X2^2, ncol = 1)

Y = apply(Y, MARGIN = 1, FUN = min, 1 )

Y = apply(matrix(Y, ncol = 1), MARGIN = 1, FUN = max, 0)

tab = as.data.frame(cbind(X1,X2,Y))

gam.regression <-gamlss(Y~X1+X2,nu.formula=~ X1+X2,tau.formula=~ X1+X2, family= BEINF, data=tab)

prob.of.0.raw <- predict(gam.regression, what="nu", type="response")

prob.of.1.raw <- predict(gam.regression, what = "tau", type = "response")

## These should be the lines you should be interested in

Do you mean RR=0 or RR=1? If RR = 0 there aren't cases of disease among the exposed to the factor, so the factor can be considered as protective. If RR=1 the incidence among the exposed to the factor is equal to the incidence among the non-exposed, that is, the incidence of the disease in the study population does not change in the presence or absence of the factor; therefore, the factor is not associated with the disease.

Hi Erol, I guess that you are looking for Analysis of Desing Basis accidents, in that case check out publlications from the UPM safety group.

In case you are looking for Severe accidents with Hydrogen, there is an excellent researcher that works with GOTHIC: Michelle Andreani from PSI, that has contributed the most to hydrogen distribution in large enclosures.

Contact me for more info.

In the USA, Regulatory Guides (RGs) 1.79 and 1.79.1 relate to the preoperational and startup testing of emergency core cooling systems (ECCS) for Pressurized Water Reactors (PWRs) and Boiling Water Reactors (BWRs). These regulatory guides identify the ECCS functions that are to be tested as those necessary to ensure the specified design functions of the ECCS are met during conditions of normal operation, anticipated operating occurrences and postulated accident conditions.

In Japan, METI (Ministry of Economy, Trade and Industry) periodic inspections are conducted every 13 months in accordance with the EUIL (Electric Utility Industry Law) for light water reactors. These inspections include ISI (in-service inspection), system performance tests, containment leakage rate tests, and overhauls of mechanical equipment. In addition, many voluntary maintenance activities, most of which are periodic overhauls for mechanical equipment, are planned and conducted by electric utilities during the plant refueling outages.

Yes, it is possible to have a negative value for lognormal mean. The main purpose of using a lognormal distribution for probabilistic analysis is to have only positive values assigned to the variables (engineering properties like hydraulic conductivity). The lognormal of the variables is normally distributed however, you would be interested in assigning the corresponding normally distributed value to the hydraulic conductivity, and in that case you will happen to raise the lognormal mean to the exponent function rendering positive values only (exponent function is always positive). Hope this helps.

check the attached file.

If your (probabilistic) model governing the data creation is well-specified, then the method of Maximum Likelihood Estimation is literally finding those parameters that optimize the likelihood function given the observed data.

Under standard assumptions, the likelihood function is the product of the probability model n times (number of data observations). This function will have specific form in the parameters.

Suggest the following to get you started:

https://books.google.at/books?id=9CaoAqaRcVwC&pg=PA281&lpg=PA281&dq=maximum+likelihood+in+risk+analysis&source=bl&ots=tPd1hDtrR4&sig=bsTgUrxFvwm6GAInFph_lrQymjY&hl=en&sa=X&ved=0ahUKEwin5Zzt0pzLAhUrkXIKHayFDlMQ6AEIMTAE#v=onepage&q=maximum%20likelihood%20in%20risk%20analysis&f=false 

Dear Anurag ,

Check the following papers

good luck

'Probabilistic analysis' is a rather wide definition of what you want to do. Why do you need to do it? What is your goal? What do you want to analyze? If you can clarify that it will be easier to come up with suitable references, because there are many of them and there is not a cookbook for it which you can just apply (actually, just applying, without considering the specific issues of your problem is a dangerous business).

As long as the different models are similar in functions, it is more efficient to maintain a single FMEA to cover them. Within the single FMEA, it is possible to denote some failure modes as specific only to Model X.  Or you could add an extra column to input which are the applicable equipment for each line item.

Hi Alberto,

I wouldn't say that European sports leagues are corruption-free as there is evidence of match fixing in lower leagues let alone some major scandals in higher leagues (example Serie A a few years ago).

However there are a number of papers that develop models which beat the betting market odds. Some very good papers are in the following journals

Regards,

Dominic

Hi,

Based on the fatigue assessment approach, uncertainties can rise from various sources.

For eg: if you chose to use S-N approach for fatigue assessment then uncertainties can arise from:

1. S-N curve

2. Miner's Rule (Linear Damage Rule) 

to account for these uncertainties we use probabilistic approach.

The Delphi technique is a widely used and accepted method for gathering expert data, to combine opinions on expected development time about a particular technology, as well as to discover what is actually known or not known about a specific topic.

In the field of radiation therapy services, an approach similar to the Delphi methodology is proposed in [1, 2]. In these papers the failure modes and human errors are initially conceived by members of the working group on an individual and independent basis (i.e. in “blind” mode), then are collectively revised during a dedicated plenary session to reach general consensus.

However, in my opinion, the following problems are not solved:

- It is usually difficult for the experts to provide advice for example in the risk data such as the occurrence probability (i.e. to give a precise answer );

- In a comparative study performed by different groups working on the same topic but in different company contexts, the investigators found only few overlap of failure modes and human errors that were identified.

1. Ciocca M, Cantone M C, Veronese I, Cattani F, Pedroli G, Molinelli S, Vitolo V, Orecchia R 2012 Application of failure mode and effects analysis to intraoperative radiation therapy using mobile electron linear accelerators Int J Radiat Oncol Biol Phys 82 (2) 305-11.

2. Cantone M C, Ciocca M, Dionisi F, Fossati P, Lorentini S, Krengli M, Molinelli S, Orecchia R, Schwarz M, Veronese I, Vitolo V 2013 Application of failure mode and effects analysis to treatment planning in scanned proton beam radiotherapy Radiation Oncology 8:127.

dear Ivan i send you one of my paper about risk management. i think that you can find something of interesting

Dieticians Dentist       DI percent DE percent                     ARR    RRR               

total=94       total=123

Chocolate   81            120          0.86            0.98              -11.39     -13.22

Yogurt        23            56            0.24            0.46              -21.06       -86.07

Fruits          67             60           0.71             0.49             22.50        31.56

Sweeties    93            123         0.99            1.00              -1.06         -1.08

First find RR, RR =0.88 for chocolates indicates dieticians are (1-0.88)% i.e 12 % percent less likely to consume chocolate compared to dentists.

The ARR is interpreted as the risk of developing the disease is increased by ARR for those individuals exposed to the risk factor. ARR of chocolate is -11.39%, thus if a person is  dietician, the  risk of chocolate consumption  reduced by 11.39% compared to dentists.

RRR measures what proportion of the risk in exposed persons is due to the exposure?

RRR of chocolates is -13.22 indicates that 13.22 % of the chocolate consumption among dietician may attributed to dieticians.

Thank you @Derek. another issue i met was with power calculation. The power calculation requires the risk allele frequency. Sometimes it is said minor allele frequency. This is not an issue under normal conditions because, most cases the risk allele will be the polymorphic allele and not wild allele. But in my case its not same, so I am confused if i should use the risk allele (long allele) or the minor allele (wild allele).

My colleague has provided a structured procedure for systematic sensitivity/uncertainty analysis in waste LCA. We now follow it in most of our LCA studies. I think it is applicable in other types of LCA as well.