Difference between revisions of "Uncertainty view of a result"

m
 
(16 intermediate revisions by 2 users not shown)
Line 2: Line 2:
 
<breadcrumbs>Analytica User Guide > Result Tables and Graphs> {{PAGENAME}}</breadcrumbs>
 
<breadcrumbs>Analytica User Guide > Result Tables and Graphs> {{PAGENAME}}</breadcrumbs>
  
Every variable has a certain or deterministic value, which we term its '''''mid '''''value. Some variables, notably chance variables and variables that depend on chance variables, can also have an uncertain or probabilistic value, which we term its '''''prob '''''value. A mid value is computed using the mid value of each variable it depends on or the median of any probability distribution. The mid value of a result is not necessarily the median of its probability distribution, but usually close.
 
  
The '''Result '''window offers seven '''''uncertainty views''''', including the mid value (which is not uncertain) and six ways to display a prob value. You can select the uncertainty views from a menu in the top-left corner of a '''Result '''window. Or you can select a variable, and select an uncertainty view option from the '''Result '''menu.
+
The [[Result window]] offers an '''Uncertainty View''' popup menu in the top-left corner to let you select how to view an uncertain quantity: 
  
[[File:Chapter2 17.png]]
+
:[[File:Chapter 1.19.png]]
  
The checkmark indicates the currently selected view.
+
Press the icon to open the popup menu so you can select one of seven uncertainty views, if the quantity is uncertain:
  
Here we illustrate each uncertainty view using the chance variable, <code>Rate_of_inflation</code>, defined as a normal distribution with a mean of 2.5 and a standard deviation of 1:  
+
:[[File:Chapter 1.21.png]]
:<code>Chance Rate_of_inflation := Normal(2.5, 1)</code>
 
  
'''Mid value''': The mid value is the deterministic value, computed by using the median instead of any input prob- ability distribution. It is computed very quickly compared to uncertain values. It is the only option available for a variable that is not probabilistic.
+
The first option, '''mid: Mid value''', is the initial default. It actually ignores uncertainty and uses the median of each uncertain quantity. It lets you evaluate the model quickly. The other options  evaluate the variable (and any uncertain predecessors if necessary) as a probability distribution, using Monte Carlo simulation. For more details, see [[Uncertainty views]] in the [[User Guide]] or the [[Tutorial]]  [[Tutorial: Open a model to browse#Displaying_alternative_uncertain_views|Displaying alternative uncertain views]].  
  
[[File:Chapter2 18.png]]
+
Depending on the chosen probability mode, the [[Result window]] will display different  probability curves:
  
<tip title="Tip">A mid value is much faster to compute than a prob(abilistic) value, since it doesn’t use Monte Carlo simulation to compute a probabilistic sample. It is often useful to look first at the mid value of a variable as a quick sanity check. Then you might select an uncertainty view, which causes its prob value to be computed if it has not already been cached.</tip>
+
:[[File:Chapter 1.22.png]]
  
'''Mean value''': An estimate of the mean (or expected value) of the uncertain value, based on the random (Monte Carlo) sample.
+
You can also select these uncertainty views from the [[Result menu]] for a selected variable.
  
[[File:Chapter2 19.png]]
+
==See Also==
 
+
<div style="column-count:2;-moz-column-count:2;-webkit-column-count:2">
<tip title="Tip">The mean and the other uncertainty views below are estimates based on the Monte Carlo (or Latin hypercube) sample. The precision of these estimates depends on the sample size and the sampling method. A larger sample size gives higher precision and takes more time and memory to compute. You can '''modify the sample size ''' and sampling method in the [[Uncertainty Setup dialog]] from the '''Result '''menu.</tip>
+
* [[Result window]]
 
+
* [[Result menu]]
'''Statistics''': A table of statistics of the uncertain value, usually, the minimum, median, mean, maximum, and standard deviation, estimated from the random sample. You can select which statistics to show in the '''[[Uncertainty Setup dialog|Statistics tab]]''' of the '''Uncertainty Setup '''dialog from the '''Result '''menu.
+
* [[Uncertainty views]]
 
+
* [[Expressing Uncertainty]]
[[File:Chapter2 20.png]]
+
* Tutorial: [[Tutorial: Open a model to browse#Examining_and_changing_uncertain_input|Examining and changing uncertain input]]
 
+
* Tutorial:  [[Tutorial: Open a model to browse#Displaying_alternative_uncertain_views|Displaying alternative uncertain views]]
'''Probability bands''': An array of percentiles (fractiles) estimated from the random sample, by default the 5%, 25%, 50%, 75%, and 95%iles. You can select which percentiles to show in the '''[[Uncertainty Setup dialog|Probability Bands tab]]''' of the '''Uncertainty Setup '''dialog from the '''Result '''menu.
+
* [[Uncertainty Setup dialog]]
 
+
* [[Evaluation Modes]]
[[File:Chapter2 21.png]]
+
* [[Mid]]
 +
* [[Mean]]
 +
* [[Sample]]
 +
* [[CDF]]
 +
* [[PDF]]
 +
* [[Fractiles]]
 +
* [[Probability Distributions]]
 +
</div>
  
'''Probability density''': Select '''probability density '''to display the uncertain distribution as a probability density function (PDF).
 
  
For a probability density function, it plots values of the quantity over the X (usually horizontal) axis, and probability density on the Y (vertical axis). Probability density shows the relative probability of different values. High values show probable regions; low values show less probable regions. The peak is the mode, the most probable value. If the density is zero, it is certain that the quantity will not have values in that range.
 
 
[[File:Chapter2 22.png]]
 
 
'''Probability mass function''': If you select '''Probability density '''for a discrete variable, it displays the variable as a '''probability mass '''function (PMF) in a bar graph with the height of each bar indicating the probability of that value.
 
 
[[File:Chapter2 23.png]]
 
 
Usually, it figures out whether to use a probability density or mass function. Very rarely, you might need to tell it the domain is discrete. See [[The domain attribute and discrete variables]] [[Choosing an appropriate distribution|Is the quantity discrete or continuous?]], and [[Probability density and mass graphs]] for more.
 
 
'''Cumulative probability''': The cumulative probability distribution (CDF) plots the possible values of the uncertain quantity along the X (usually horizontal) axis. The Y value (usually height) of the graph at each value of X shows the probability that the quantity is less than or equal to that X value. The CDF must start at a probability of 0 on the extreme left and increase to a probability of 1 on the extreme right, never decreasing.
 
 
The steeper the curve, the more likely the quantity will have a value in that region. The PDF is the slope (first derivative) of the CDF. Conversely, the CDF is the cumulative integral of the PDF.
 
 
[[File:Chapter2 24.png]]
 
 
'''Sample''': A sample is an array of the random values from the distribution generated by the Monte Carlo sampling process. The sample is the underlying form used to represent each uncertain quantity. All the other uncertainty views use statistics estimated from the sample. The sample view gives more detail than you usually want. You will likely want to view it mainly when verifying or debug- ging a model.
 
 
[[File:Chapter2 25.png]]
 
 
Like any other graph, you can display a sample as a table by clicking [[File:Chapter2 4.png]] to see the underlying numerical values.
 
 
[[File:Chapter2 26.png]]
 
 
==See Also==
 
 
<footer>Graph view of a result / {{PAGENAME}} / Comparing results </footer>
 
<footer>Graph view of a result / {{PAGENAME}} / Comparing results </footer>

Latest revision as of 17:28, 12 August 2016


The Result window offers an Uncertainty View popup menu in the top-left corner to let you select how to view an uncertain quantity:

Chapter 1.19.png

Press the icon to open the popup menu so you can select one of seven uncertainty views, if the quantity is uncertain:

Chapter 1.21.png

The first option, mid: Mid value, is the initial default. It actually ignores uncertainty and uses the median of each uncertain quantity. It lets you evaluate the model quickly. The other options evaluate the variable (and any uncertain predecessors if necessary) as a probability distribution, using Monte Carlo simulation. For more details, see Uncertainty views in the User Guide or the Tutorial Displaying alternative uncertain views.

Depending on the chosen probability mode, the Result window will display different probability curves:

Chapter 1.22.png

You can also select these uncertainty views from the Result menu for a selected variable.

See Also


Comments


You are not allowed to post comments.