Capturing Uncertainty

Not all Pro and Con are certain to exist.

Say, you want to decide if you want to rent an apartment or buy a house. You have listed following Pros and Cons.

Pros Cons List

Now, think about it. There can be some expenses which will be paid by Landlord. But, how likely you will save the maintenance cost. Within a year lease, there can be 50% chance that you may face something broken which will be paid by the landlord. It may happen or it may not happen.

Also, think about the con "Could be forced to move out if the owner sells the property". It is definitely a Con when renting an apartment. If that happens, it can highly impact you, but how likely that can happen? Very low probability, right? Ok, let's think that the probability is 25%.

Also, we can assume that, after renting a property, there can be lots of restrictions, but the rental management can be very liberal too. You do not know yet. So, that is an uncertain Con as well.

When analyzing Pros and Cons, you should consider the Importance or Impact of a Pro and Con. You should also consider the Probability of an Uncertain Pro / Con. Otherwise, you may choose an option with a Good Looking Pro where the probability of the outcome of that Pro is very low. That will be a bad decision, right?

Using the SpiceLogic Pros and Cons Analyzer, you can mark a Pro/Con as Uncertain and then you can specify the Probability of that Pro/Con. Then this tool will calculate the Expected Value based on that probability in the final calculated. That will be a very reasonable way of performing the analysis.

Here is the screenshot of the Pros/Cons where Uncertainty checkbox is checked to mark a Pro/Con as Uncertain. Also using a slider, the probability of the uncertain Pro/Con is specified.

Capture_ Uncertainty

Risk Profile View

When you incorporate Uncertainty in your Pros/Cons model, this application can calculate Risk Profiles for an Option. A Risk profile is a chart that shows every possible combination of outcome along with the probability of that outcome. It is a very useful tool to understand the overall picture of an Uncertain Option. In the following chart, the X-axis shows the possible outcome, and the Y-axis shows the probability of that outcome.

Risk_ Profile

Say, you have uncertainty like an event A can occur or Not occur. Also, you may have an event B, may occur or may not occur.

So, basically, you have a combination of 4 events like this:

1. event A occurs AND event B occurs
2. event A occurs AND event B does not occur.
3. event A does not occur AND event B occurs.
4. event A does not occur AND event B does not occur.

A Risk profile chart shows net worth (outcome) of an option for all of these possible combinations along with probability. So, let's examine the above chart. We see that Buying a house option outcomes can be worth of =-15 to 25 with a probability distribution where the maximum probability of the outcome is around '5'. It also tells that, for buying a house option, the WORST CASE, you can expect to have -15, which is a NEGATIVE Utility, and BEST CASE, you can have 25 unit of worthiness which is way more than the BEST CASE of 'Renting an apartment'. Also, notice that the WORST CASE of the 'Renting an apartment' option is around -6. Which is better than the WORST CASE of Buying a house. So, if you choose to rent an apartment, you have less risk to worry about, comparing the option 'Buying a house'.

Cumulative distribution function and Survival Function

When you have uncertainty in your decision model, two additional views can be generated to better understand the nature of uncertainty. They are 'Cumulative distribution function' and 'Survival function'. It is not very necessary to examine these 2 charts for simple Pros and Cons problems, but it can be interesting for many students or analyst, and therefore, the application offers these features.

cdf_and_sf_button

Value Set

The Risk Profile Chart also shows a tab named 'Value Set'. The value set is simply a chart showing the values of all possible outcomes, small to big order without any probability information. This chart is useful to understand, the WORST CASE, BASE CASE, RANGE of POSSIBLE OUTCOME in a straightforward order, without the complexity of Probability. For example, here is a sample ValueSet chart for a problem 'Should I take the job ?'.

value_set

Here, you can understand that, the option has Minimum worthiness value = -55 and Maximum worthiness value = 25. It has some intermediate worthiness value as 45 and 5. This chart won't give you the idea of what outcome may occur at what probability, but it can give you a clear view of what are the possibilities.

Most Likelihood Value

When you model uncertainty in a decision context, you will see a Metric chooser section where you can choose to show Most Likelihood Value.

most_likelihood_value

Most likelihood value is the value that gives you the calculated worthiness of an option, considering only those outcomes that have the highest probabilities.

For example, let's consider the same set of the event as we discussed earlier:

An event A can occur (with probability 25%) or Not occur (with probability 75%).
Also, you may have an event B, may occur (with probability 60%) or may not occur (with probability 40%)

The most likelihood strategy is that only the highest probability outcome event should be counted and the rest of the events will be discarded from the calculation.

So, there, we see the event A, the probability of NOT to occur is higher than the probability to occur. So, the strategy is to consider that, event A will not occur. So, it will assign 0 unit to Event A.

For event B, the probability to occur is higher than not to occur. So, it will consider that event B will occur and count the worthiness of event B.

Finally, the chart will show the total worthiness of an option based on the strategy mentioned above.

It is not always practical to depend on Most Likelihood Value all the time. Because, in the real world, you do not always count an event which has the highest probabilities. You may need to consider the other outcomes that have lower probabilities, right?

Therefore, it is always recommended to be concerned with the 'Expected Utility' chart rather than 'Most likelihood Value' chart.

Last updated on 21 July 2018, Saturday, 10:45:03 PM
If you have any questions or concerns about this tutorial, Please feel free to share your comment.