Discussion 1:Daniela

Based on the reading I have done on the issue, I will disagree with the argument against assuming distributions. Here is the reason why: assuming parameter values feels like a deterministic method of modeling, which we know it not the best way of assessing uncertainty, which is the essence of risk. Assigning probabilities of random input values takes into consideration more than one scenario, inherently includes change of outcome, and results in a range of possible outcomes, as opposed to only one. This is a more realistic way of relating to uncertainty. (Palisade).

Risk Analysis. Palisade. https://www.palisade.com/risk/risk_analysis.asp

Discussion 2: Tyler

The distribution of a risk analysis simulation is very important to accurately and precisely predict the uncertainty, randomness, and variability of the specific problem (Vose Software, 2017). Vose software also claims from their experience that inappropriate use of probability distributions have proven to be a very common failure of risk analysis models. This is largely because this represents an “inadequate understanding of the theory behind probability distribution functions and, in part, from failing to appreciate the knock-on effects of using inappropriate distributions” (Vose Software, 2017).

Several properties of your data must be understood to determine what type of distribution to use in a simulated model. You would want to determine if your data is discrete or continuous, bounded or unbounded, parametric or non-parametric, univariate or multivariate, and if it is first or second order. The reference I have to Vose Software below is incredibly valuable as it provides assistance to determine what kind of distribution should be specified according to your data and the properties is has related to the properties above. Assuming parameters can be very risky and one should definitely spend the time to make sure that you are using the correct distribution to avoid creating a model that will ultimately fail you or your company.

References

Vose Software (2017). Selecting the appropriate distributions for your model. Vose Software. Retrieved March 24, 2022, from https://www.vosesoftware.com/riskwiki/Selectingtheappropriatedistributionsforyourmodel.php

Discussion 3: Arcelia

As discussed by Yse (2020), probability is focused on analyzing how random events will occur. An assumed parameter is not random and therefore cannot provide us with an accurate representation of real-life events. As we learned last week, assumed parameter values that are not truly “random” will cause an outcome of a simulation to fall prey to the flaw of averages.

Specifically, a probability distribution “is a list of all the possible outcomes of a random variable, along with its corresponding probability values” (Yse, 2020, Probability Distributions section). Probability distribution models, as discussed by Albright & Winston (2017), allow you to select values and their probabilities that fit the data being modeled. Therefore, you must understand your data and choose the distribution that best meets the needs of your analysis. For example, will your random variable need to be discrete or continuous? The answer to that question allows you to choose the type of distribution model to use (a normal or triangular distribution if the random variable is continuous, or a discrete or binomial distribution if it is discrete, etc.).

Ultimately, as noted by Albright & Winston (2017), the outcome of a simulation is influenced by the type of distribution used. Using, and knowing when to utilize distributions allows us to model real-life scenarios which in turn provides us with meaningful estimates that account for variability (Yse, 2020).

References

Albright, S. C., & Winston, W. L. (2017) Business

analytics: Data analysis & decision making (6th ed.). Cengage. ISBN-13: 9781305947542.

Yse, D. L. (2020, March 16). Before Probability Distributions – Towards Data Science. Towards Data Science. https://towardsdatascience.com/before-probability-distributions-d8a2f36b1cb#:%7E:text=Probability%20distributions%20help%20to%20model,the%20probability%20of%20an%20event.