Tutorials › Exam P - Probability › Conditional Distribution and Independence Problem Workshop

Home  /  Tutorials  /  Exam P / Conditional Distributions and Independence

Exam P · Conditional Distributions and Independence · Tutorial 752 of 915

Conditional Distribution and Independence Problem Workshop

Build reliable solutions to conditional-distribution problems by tracking supports, normalizing correctly, and checking independence with joint and marginal laws.

Intermediate 10 min read

What You'll Learn

  • Normalize discrete conditional probabilities using the probability of the conditioning category.
  • Derive conditional densities from joint and marginal densities.
  • Restrict integration to the joint support and the event being evaluated.
  • Compute conditional probabilities for events involving intervals or regions.
  • Test independence using joint-marginal factorization and conditional laws.
  • Recognize denominator and support errors that can change an answer.

Solving Conditional-Distribution Problems Systematically

Conditional-distribution problems often combine several tasks: identify which values are possible after conditioning, normalize a conditional law, calculate an event probability, and decide whether the variables are independent. The central discipline is to keep the conditioning event and the allowed support visible throughout the calculation. A correct formula applied over the wrong range, or divided by the wrong marginal probability, gives an incorrect result.

For a discrete pair, conditioning means dividing joint probabilities by the probability of the observed value of the conditioning variable. For a continuous pair, conditional densities are obtained by dividing the joint density by the appropriate marginal density, where that marginal is positive. In either setting, event probabilities must use only outcomes allowed by the joint model.

Definition: For discrete random variables \(X\) and \(Y\), if \(P(X=x)>0\), then $$ P(Y=y\mid X=x)=\frac{P(X=x,Y=y)}{P(X=x)}. $$ For continuous random variables with joint density \(f_{X,Y}\), if \(f_X(x)>0\), a conditional density of \(Y\) given \(X=x\) is $$ f_{Y\mid X}(y\mid x)=\frac{f_{X,Y}(x,y)}{f_X(x)}. $$ The conditional density is zero outside the slice of the joint support corresponding to that \(x\). For continuous variables, this formula describes conditioning through a conditional density; it does not treat the probability of the event \(X=x\) as positive.

A Reliable Problem-Solving Sequence

Before calculating, identify the conditioning variable and value or event. Then find its probability or marginal density. Next, restrict the numerator or integral to the outcomes that satisfy both the joint support and the event of interest. Finally, normalize and check that the result is a valid probability or density.

1
Write down the support.
Record the possible pairs \((x,y)\) before choosing any summation indices or integration bounds.
2
Find the conditioning denominator.
For discrete conditioning, use the appropriate marginal probability. For continuous conditioning, use the marginal density at the specified value, when it is positive.
3
Restrict to the requested event.
Intersect the event with the support; do not include points where the joint probability or density is zero.
4
Check the result.
Conditional probabilities must lie between 0 and 1, and a conditional density must integrate to 1 over its conditional support.

The conditional-density formula has a direct justification. For a short interval of \(X\)-values near \(x\) and a set \(B\) of possible \(Y\)-values, the joint probability of both events is approximately the joint density near \(x\), integrated over \(B\), times the interval width. The probability of the \(X\)-interval is approximately \(f_X(x)\) times that width. Dividing cancels the width, leaving the joint density divided by \(f_X(x)\), integrated over \(B\). Thus, for suitable sets \(B\),

$$ P(Y\in B\mid X=x)=\int_B \frac{f_{X,Y}(x,y)}{f_X(x)}\,dy. $$

This is why the conditional density is a ratio of densities, rather than a joint density used without normalization. It also explains why its support can depend on the conditioning value.

Worked Example: Conditional Probabilities From a Discrete Table

A hypothetical inspection process records a component’s category \(X\), which is A, B, or C, and whether it passes a secondary check, recorded as \(Y=1\) for pass and \(Y=0\) otherwise. The joint probabilities are shown below. Find \(P(Y=1\mid X=A)\), \(P(X=A\mid Y=1)\), and \(P(X\ne B\mid Y=1)\). Then test whether \(X\) and \(Y\) are independent.

\(Y=0\)\(Y=1\)Marginal \(P(X=x)\)
\(X=A\)0.100.200.30
\(X=B\)0.200.100.30
\(X=C\)0.100.300.40
Marginal \(P(Y=y)\)0.400.601.00

Method. Use the conditional-probability ratio, with the marginal probability of the conditioning event in the denominator. The table entries sum to \(0.10+0.20+0.20+0.10+0.10+0.30=1.00\), as required.

For the first conditional probability, the conditioning event is \(X=A\), whose probability is \(0.30\). Therefore,

$$ P(Y=1\mid X=A)=\frac{P(X=A,Y=1)}{P(X=A)} =\frac{0.20}{0.30}=\frac{2}{3}\approx 0.6667. $$

For the second, the conditioning event is \(Y=1\), whose marginal probability is \(0.60\):

$$ P(X=A\mid Y=1)=\frac{0.20}{0.60}=\frac{1}{3}\approx 0.3333. $$

The event \(X\ne B\), given \(Y=1\), includes the two disjoint cells \((A,1)\) and \((C,1)\). Thus,

$$ P(X\ne B\mid Y=1)=\frac{0.20+0.30}{0.60} =\frac{0.50}{0.60}=\frac{5}{6}\approx 0.8333. $$

To test independence, compare a joint probability with the product of its marginals. If \(X\) and \(Y\) were independent, \(P(X=A,Y=1)\) would equal \(P(X=A)P(Y=1)=0.30(0.60)=0.18\). The actual joint probability is \(0.20\), so the variables are dependent. Each conditional probability is between 0 and 1, and each division uses the marginal of the stated conditioning event.

Common mistake avoided: For \(P(X=A\mid Y=1)\), dividing by \(P(X=A)=0.30\) would reverse the conditioning. The denominator must be \(P(Y=1)=0.60\).

Continuous Supports and Conditional Events

For continuous models, the joint support determines the limits of integration. An event such as \(Y>c\) may cut through the support, so the allowed range of \(Y\) can change with \(X\). When conditioning on an event such as \(X>a\), the conditional probability is the probability of the intersection divided by the probability of the conditioning event:

$$ P(A\mid B)=\frac{P(A\cap B)}{P(B)},\qquad P(B)>0. $$

In a joint-density calculation, the numerator integrates over the part of the joint support satisfying both \(A\) and \(B\). This remains the appropriate approach even when an event boundary divides the support into pieces.

Worked Example: A Conditional Event on a Triangular Support

Let \((X,Y)\) have joint density \(f_{X,Y}(x,y)=2\) for \(0<x<y<1\), and zero elsewhere. Find \(P(Y>3/4\mid X>1/2)\). Also determine whether \(X\) and \(Y\) are independent.

Method. First obtain the marginal density of \(X\), then compute the conditioning probability and the intersection probability using the triangular support. The density is valid because the area of \(0<x<y<1\) is \(1/2\), and \(2(1/2)=1\).

For a fixed \(x\), the support requires \(x<y<1\). Hence

$$ f_X(x)=\int_x^1 2\,dy=2(1-x),\qquad 0<x<1. $$

The conditioning probability is

$$ P(X>1/2)=\int_{1/2}^1 2(1-x)\,dx =[2x-x^2]_{1/2}^{1} =1-\frac{3}{4}=\frac{1}{4}. $$

For the numerator, impose both \(x>1/2\) and \(y>3/4\), while retaining \(x<y<1\). When \(1/2<x<3/4\), \(y\) ranges from \(3/4\) to 1. When \(3/4\le x<1\), the support already requires \(y>x\). Therefore,

$$ \begin{aligned} P(Y>3/4,\ X>1/2) &=\int_{1/2}^{3/4}\int_{3/4}^1 2\,dy\,dx +\int_{3/4}^1\int_x^1 2\,dy\,dx\\ &=2\left(\frac14\right)\left(\frac14\right) +\int_{3/4}^1 2(1-x)\,dx\\ &=\frac18+\left[2x-x^2\right]_{3/4}^1\\ &=\frac18+\left(1-\frac{15}{16}\right) =\frac18+\frac{1}{16}=\frac{3}{16}. \end{aligned} $$

Thus,

$$ P(Y>3/4\mid X>1/2)=\frac{3/16}{1/4}=\frac34=0.7500. $$

For independence, the other marginal density is \(f_Y(y)=\int_0^y 2\,dx=2y\) for \(0<y<1\). On the open region \(0<y<x<1\), which has positive area, the joint density is zero, but \(f_X(x)f_Y(y)=2(1-x)\,2y>0\). The densities therefore fail to factor on a region of positive area, so \(X\) and \(Y\) are dependent.

Common mistake avoided: Integrating over all \(1/2<x<1,\ 3/4<y<1\) as a rectangle would include points with \(y\le x\), which are outside the support. The split at \(x=3/4\) accounts for the changing lower bound.

Factorization and Conditional-Law Checks

For discrete variables, independence means \(P(X=x,Y=y)=P(X=x)P(Y=y)\) for every possible pair. For continuous variables, the corresponding requirement is \(f_{X,Y}(x,y)=f_X(x)f_Y(y)\) almost everywhere. The “almost everywhere” qualification matters: changing a density at an isolated point does not change any probability. A mismatch over a region of positive area, however, rules out independence.

A second useful check is to compare conditional laws. Under independence, conditioning on a value of one variable does not change the distribution of the other, wherever the relevant conditional law is defined. If the conditional probabilities or densities vary with the conditioning value, that is evidence of dependence. Conversely, an independence conclusion should be based on a valid factorization or a justified equality of conditional laws, not merely on one numerical coincidence.

Worked Example: Product Density and Conditional Event Probabilities

Suppose \(X\) and \(Y\) have joint density \(f_{X,Y}(x,y)=4xy\) on the square \(0<x<1,\ 0<y<1\), and zero elsewhere. Find \(P(Y>1/2\mid X>1/2)\), calculate \(P(X<Y)\), and determine whether the variables are independent.

Method. Integrate the joint density to obtain the marginals, then check whether the joint density factors. For a fixed \(x\),

$$ f_X(x)=\int_0^1 4xy\,dy=4x\left[\frac{y^2}{2}\right]_0^1=2x, \qquad 0<x<1. $$

Similarly, \(f_Y(y)=\int_0^1 4xy\,dx=2y\) for \(0<y<1\). Both integrate to 1, and their product is \(f_X(x)f_Y(y)=(2x)(2y)=4xy\), which equals the joint density throughout the square. Thus \(X\) and \(Y\) are independent.

Independence gives \(P(Y>1/2\mid X>1/2)=P(Y>1/2)\). From the marginal density of \(Y\),

$$ P(Y>1/2)=\int_{1/2}^1 2y\,dy=[y^2]_{1/2}^1 =1-\frac14=\frac34=0.7500. $$

For the event \(X<Y\), integrate over the triangular region \(0<x<y<1\):

$$ \begin{aligned} P(X<Y) &=\int_0^1\int_x^1 4xy\,dy\,dx\\ &=\int_0^1 4x\left[\frac{y^2}{2}\right]_x^1 dx =\int_0^1 (2x-2x^3)\,dx\\ &=[x^2-\tfrac12 x^4]_0^1 =1-\frac12=\frac12=0.5000. \end{aligned} $$

The event probability lies between 0 and 1, and the marginal densities each integrate to 1. Common mistake avoided: A product density does not make every event probability a product. For \(X<Y\), the event restricts the pair to a triangle, so the joint density must be integrated over that triangle.

Common Traps and Final Checks

  • Using the wrong denominator. In \(P(A\mid B)\), divide by \(P(B)\), not by \(P(A)\). For discrete variables, identify the conditioning row or column before dividing.
  • Forgetting a restricted support. A rectangular-looking event may intersect a triangular support only in part. Draw or describe the intersection before setting bounds.
  • Assuming independence from one matching value. Independence requires factorization for all discrete pairs, or almost everywhere for densities. One matching cell or point does not establish it.
  • Claiming dependence from an isolated mismatch. For continuous densities, a difference at one point is not enough; the mismatch must persist on a set of positive area to affect the density factorization test.
  • Failing to normalize. A conditional density must integrate to 1 over its conditional support. If it does not, recheck the marginal denominator and bounds.

A useful final audit is to state the support, state the conditioning event, and verify the denominator before simplifying. This combination of conditional laws, event regions, and factorization checks is aligned with the SOA Exam P (Probability) syllabus.

Check Your Understanding

For each problem, write the conditioning denominator and the relevant support or table entries before calculating.

  1. A discrete pair has joint probabilities \(P(X=M,Y=0)=0.12\), \(P(X=M,Y=1)=0.18\), \(P(X=N,Y=0)=0.28\), and \(P(X=N,Y=1)=0.42\). Find \(P(Y=1\mid X=N)\) and \(P(X=M\mid Y=1)\), and decide whether \(X\) and \(Y\) are independent. (Answer: \(0.6000;\ 0.3000;\) independent.)
  2. Let \(f_{X,Y}(x,y)=2\) on \(0<x<y<1\), and zero elsewhere. Find \(P(Y>0.8\mid X>0.6)\). (Answer: \(0.7500\).)
  3. Let \(f_{X,Y}(x,y)=4xy\) on \(0<x<1,\ 0<y<1\), and zero elsewhere. Find \(P(X<0.5\mid Y>0.5)\), and decide whether the variables are independent. (Answer: \(0.2500;\) independent.)
  4. Let \(f_{X,Y}(x,y)=2\) on \(0<y<x<1\), and zero elsewhere. Find \(P(Y<0.5\mid X>0.5)\), and determine whether the variables are independent. (Answer: \(0.6667;\) dependent.)
  5. A discrete pair has \(P(X=1,Y=1)=0.15\), \(P(X=1,Y=0)=0.35\), \(P(X=0,Y=1)=0.25\), and \(P(X=0,Y=0)=0.25\). Find \(P(X=1\mid Y=1)\) and decide whether the variables are independent. (Answer: \(0.3750;\) dependent.)

Previous

Next

Back to Tutorials
Clinical Biostats