The Secret Inside Every Average
Who gets to influence an average, and how does that become the Law of Total Probability?
How this lesson unfolds.
Begin with What is the hidden question inside an average? The sequence then connects Equal averaging assumes equal influence and builds toward Recognize total probability before calculating.
What is the hidden question inside an average?
An average is a rule for contribution. It combines values and decides how much influence each value receives.
- Learning goalExplain average as contribution and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionEqual averaging assumes equal influence
When does the ordinary mean make sense?
The arithmetic mean assigns every observation weight 1/n. It is appropriate when each listed observation is intended to contribute equally.
- Learning goalExplain equal weight mean and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionEqual contribution is not always fair
What changes when observations represent different amounts of reality?
If groups differ in size, reliability or frequency, equal weights can misrepresent the population. Weights should match the contribution the problem defines.
If groups differ in size, reliability or frequency, equal weights can misrepresent the population.
- Learning goalExplain weight choice and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionEach world has its own local probability
How can one question have several correct local answers?
A conditional probability answers the question inside one world or condition. Different worlds can therefore have different correct values for the same event.
- Learning goalExplain conditional worlds and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionNature chooses a world before the outcome
What happens before the event we observe?
A partition models mutually exclusive worlds. Nature first lands in one world Hᵢ with probability P(Hᵢ), then the event A occurs with local probability P(A|Hᵢ).
- Learning goalExplain generative story and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionA probability tree exposes the weighted average
How do branch probabilities combine?
Multiply probabilities along a branch to obtain P(A∩Hᵢ), then add across disjoint branches that lead to A.
- Learning goalExplain tree computation and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionThe Law of Total Probability writes itself
What is the general formula?
For a partition H₁,…,Hₖ, total probability is P(A)=ΣᵢP(A|Hᵢ)P(Hᵢ). It is a weighted average of local probabilities.
For a partition H₁,…,Hₖ, total probability is P(A)=ΣᵢP(A|Hᵢ)P(Hᵢ).
- Learning goalExplain total probability formula and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionWeights measure influence, not value
Why is the overall pass rate 63 percent rather than 75 percent?
If 10% of students attend a 90%-pass school and 90% attend a 60%-pass school, the overall rate is 0.1×0.9+0.9×0.6=0.63. The larger group has more influence.
If 10% of students attend a 90%-pass school and 90% attend a 60%-pass school, the overall rate is 0.1×0.9+0.9×0.6=0.63.
- Learning goalExplain weighted example and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionRecognize total probability before calculating
How can you spot this structure in a new problem?
Look for hidden histories or cases that are mutually exclusive and exhaustive, local conditional probabilities, and known case weights. Then multiply within cases and add across cases.
- Learning goalExplain recognition framework and choose justified weights.
- Helpful foundationsarithmetic mean · conditional probability · probability trees
- Common trapConditional probabilities must be weighted by the probabilities of their conditioning cases; they are not generally averaged equally.
- Next connectionUse the same mixture logic for total expectation