EE8405 Chap.1 Probability and Conditional Reasoning
Probability and Conditional Reasoning
Conditional reasoning supplies the language used to describe uncertain arrivals, errors and observations. The chapter separates events from random variables, builds conditional probability from joint probability, reverses conditions with Bayes' rule and tests independence without confusing it with mutual exclusivity.
Its worked examples keep base rates visible so a high detector sensitivity is not mistaken for a high posterior probability.
What this chapter covers
- 01
Sample spaces and events
- 02
Conditional probability
- 03
Bayes' rule
- 04
Total probability
- 05
Independence tests
- 06
Base-rate effects
Worked example · free
Reverse an alarm probability
- 2Compute each joint alarm route.
- 2Add routes for the evidence total.
- 2Divide and interpret the posterior.
Key terms
- Conditional Probability
- Probability restricted to outcomes where the stated evidence occurs.
- Independence
- A relation in which learning one event does not change the probability of another.
Probability and Conditional Reasoning FAQ
Why can a sensitive alarm have a low posterior?
A rare target can generate fewer true alarms than the common non-target population generates false alarms. Bayes' rule retains both routes in the evidence denominator, so the prior remains decisive.
Are mutually exclusive events independent?
Nonempty mutually exclusive events are not independent because observing one makes the other impossible. Independence instead requires the joint probability to equal the product of the marginals.
What is the safest Bayes workflow?
Define events, state the requested direction, multiply likelihoods by priors, sum all routes producing the evidence, and divide the target route by that evidence total before interpreting the answer.
How do posterior odds provide a second check?
Convert the prior probability into odds, multiply by the likelihood ratio for the observed evidence, and convert the resulting posterior odds back to probability. This calculation should agree with the probability-tree answer apart from rounding. If it does not, inspect the direction of the likelihood ratio and whether every evidence-producing route was included in the tree denominator.
The check is independent enough to expose a reversed conditional probability.
Why does the reference population matter in conditional reasoning?
Detector characteristics and priors belong to the population and time period in which they were measured. Applying a quiet-period false-alarm rate during peak traffic or combining a prior from one customer group with a likelihood from another produces a polished but unsupported posterior.
State the reference class, identify any transport assumption and explain what evidence would show the rates remain stable in the application population.
What must the evidence denominator contain?
Include every mutually exclusive route that produces the observation. Missing one understates the denominator; overlapping routes count the same outcome twice and distort the posterior.
Exam move
Draw probability trees and two-by-two tables for the same problems. If both routes produce the same posterior, the conditioning direction and denominator are probably correct. Define the experiment, sample space and events before attaching numbers.
Check that the branches forming the denominator are mutually exclusive and exhaustive, because a missing route understates the evidence probability while overlap counts outcomes twice. Write the requested conditional probability in words and symbols, paying attention to which event is after the conditioning bar.
For Bayes problems, calculate every joint route that produces the evidence, add those routes, and divide the target route by the evidence total. Interpret the result against the prior so a high likelihood is not mistaken for a high posterior. Use posterior odds as an independent check: multiply prior odds by the likelihood ratio and convert back to probability. The two methods should agree apart from rounding.
Practise rare-event cases because they make base-rate errors obvious. Change only the prior while keeping detector properties fixed and predict the direction of the posterior before recalculating. Separate independence from mutual exclusivity. Verify independence with the product rule or by showing that conditioning leaves a probability unchanged.
For two nonempty mutually exclusive events, observing one makes the other impossible, so they cannot be independent. Include examples from telecommunications: independent source arrivals may overlap, whereas distinct channel states at one instant are mutually exclusive. Preserve the reference population whenever a probability comes from data.
A false-alarm rate measured during quiet periods may not transfer to peak traffic, and a correct fraction with mismatched populations is not a correct inference. State what evidence would support stable rates. Keep one two-by-two table template on the handwritten page with cells for true event, non-event, evidence and non-evidence. Add the total-probability denominator and an odds check.
In timed work, finish by explaining what changed after the observation, why the prior mattered, and whether the probabilities belong to the application population.
Working through Probability and Conditional Reasoning in EE8405? Sia is AskSia’s AI Electrical Engineering tutor — ask any EE8405 Probability and Conditional Reasoning question and get a clear, step-by-step explanation grounded in how EE8405 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
