What the Joint CDF Measures
When two random variables describe the same outcome, a probability question can place a threshold on each variable at once. For example, a service center might ask for the probability that processing takes at most two days and that the case is not escalated. The joint cumulative distribution function answers this kind of question for any pair of thresholds.
The event in this definition consists of outcomes whose coordinates satisfy \(X\le x\) and \(Y\le y\). On a graph with \(X\) on the horizontal axis and \(Y\) on the vertical axis, those outcomes lie in the rectangle extending left from \(x\) and down from \(y\). This is called a lower-left rectangle. The boundaries at \(x\) and \(y\) are included because the definition uses “less than or equal to.”
The joint CDF is a function of two real thresholds. Its value is a probability, so it must lie between zero and one. Increasing either threshold cannot decrease the value: a larger threshold includes every outcome counted at the smaller threshold, and possibly more. These basic checks are useful when evaluating a proposed formula or calculation.
Calculating from a Discrete Joint Distribution
For discrete random variables, the joint distribution assigns probability to possible pairs of values. To find \(F_{X,Y}(x,y)\), include every pair \((a,b)\) for which \(a\le x\) and \(b\le y\), then add the probabilities of those pairs. The thresholds need not themselves be possible values of either random variable.
Worked Example: Reading a Joint CDF from a Probability Table
A hypothetical monitoring system records \(X\), the number of alerts in an hour, and \(Y\), an indicator equal to 1 if a review is required and 0 otherwise. The joint probability table is:
| \(X\) | \(P(X=x,Y=0)\) | \(P(X=x,Y=1)\) |
|---|---|---|
| 0 | 0.10 | 0.15 |
| 1 | 0.20 | 0.05 |
| 2 | 0.25 | 0.25 |
Find \(F_{X,Y}(1,0)\), \(F_{X,Y}(1.5,0.5)\), and \(F_{X,Y}(1,1)\). The method is to identify the table entries satisfying both threshold conditions and sum their probabilities.
First, \(F_{X,Y}(1,0)\). The conditions are \(X\le1\) and \(Y\le0\). Since \(Y\) can only be 0 or 1, only the \(Y=0\) column qualifies. The qualifying \(X\) values are 0 and 1:
Next, \(F_{X,Y}(1.5,0.5)\). The condition \(X\le1.5\) includes \(X=0\) and \(X=1\), while \(Y\le0.5\) includes only \(Y=0\). Thus:
Finally, \(F_{X,Y}(1,1)\). Both possible values of \(Y\) satisfy \(Y\le1\), and \(X=0,1\) satisfy \(X\le1\). Therefore:
As a check, the six table entries sum to \(0.10+0.15+0.20+0.05+0.25+0.25=1.00\), and each requested CDF value is between zero and one. The second result matches the first because the thresholds \(1.5\) and \(0.5\) select the same possible values as 1 and 0.
Why the CDF Formula Works
The definition follows directly from the event being measured. For each possible outcome, ask whether its \(X\)-coordinate is at most \(x\) and its \(Y\)-coordinate is at most \(y\). If both statements are true, the outcome belongs to the lower-left rectangle and contributes its probability. If either statement is false, it contributes nothing. For a discrete joint distribution, this gives the finite or countable sum
where the sum runs over the possible pairs of values. This is not a separate probability rule; it is the definition applied by partitioning the event into disjoint possible pairs. The same event-based definition applies when the variables are continuous, although probabilities then need not be obtained by summing point probabilities.
One consequence is monotonicity. If \(x_1\le x_2\) and \(y_1\le y_2\), every outcome satisfying \(X\le x_1\) and \(Y\le y_1\) also satisfies \(X\le x_2\) and \(Y\le y_2\). The first event is contained in the second, so its probability cannot be larger:
Inclusive Boundaries Matter
The weak inequalities in the definition are essential for discrete variables. If a variable can equal a threshold with positive probability, outcomes on that boundary are included. A CDF value is a probability of being at or below both thresholds; it is not the probability of taking exactly the two threshold values. The following example makes the distinction explicit.
Worked Example: Applying the Thresholds to a Discrete Scenario
In a hypothetical document-processing study, \(X\) is the number of days to complete a file and \(Y\) equals 1 if the file is escalated, or 0 otherwise. The possible pairs and probabilities are:
| \((X,Y)\) | Probability |
|---|---|
| \((1,0)\) | 0.20 |
| \((2,0)\) | 0.15 |
| \((2,1)\) | 0.10 |
| \((3,0)\) | 0.25 |
| \((4,1)\) | 0.20 |
| \((5,1)\) | 0.10 |
Find the probability that completion takes no more than two days and escalation occurs, and then find \(F_{X,Y}(2,1)\). The six probabilities sum to \(0.20+0.15+0.10+0.25+0.20+0.10=1.00\), so the stated probabilities form a valid distribution.
Probability of no more than two days and escalation. This event requires \(X\le2\) and \(Y=1\). Among the listed pairs, only \((2,1)\) qualifies, so the probability is \(0.10\).
Joint CDF value \(F_{X,Y}(2,1)\). Here the conditions are \(X\le2\) and \(Y\le1\). Because \(Y\) takes only 0 and 1, both values qualify. The qualifying pairs are \((1,0)\), \((2,0)\), and \((2,1)\):
The two answers differ because the first event requires \(Y=1\), whereas the CDF event includes both \(Y=0\) and \(Y=1\). The CDF answer also includes \(X=2\), since equality at the threshold is allowed.
A Continuous Pair: Probability from the Event
For a continuous pair, a joint CDF value is still found by identifying the outcomes in the lower-left rectangle. One way to calculate it is to represent both variables using a single underlying random quantity and determine which values of that quantity satisfy both inequalities. This avoids treating individual points as if they had positive probability.
Worked Example: A Joint CDF for Two Related Uniform Variables
Let \(U\sim\operatorname{Uniform}(0,1)\), and define \(X=U\) and \(Y=1-U\). Thus \(X\) and \(Y\) are both between 0 and 1, and their values are linked by \(X+Y=1\). Find \(F_{X,Y}(0.7,0.6)\) and \(F_{X,Y}(0.4,0.6)\).
The method is to translate both CDF conditions into restrictions on \(U\). For the first pair of thresholds:
Both conditions hold when \(0.4\le U\le0.7\). A uniform random variable on \([0,1]\) assigns probability equal to interval length, so:
For the second pair of thresholds, \(X\le0.4\) requires \(U\le0.4\), while \(Y\le0.6\) still requires \(U\ge0.4\). Their intersection contains only \(U=0.4\). A continuous uniform random variable assigns probability zero to a single point, giving:
The answers are consistent with the support relation \(X+Y=1\): the first lower-left rectangle intersects a segment of possible outcomes, while the second touches that segment only at one point. The interval length calculation is exact, so no rounding is needed.
Common Mistake: Replacing “At Most” with “Less Than”
A frequent error is to exclude values equal to a threshold even when the problem asks for a CDF value. In the first table example, an incorrect calculation of \(F_{X,Y}(1,0)\) might use only \(X=0\), reasoning that \(X\) must be strictly below 1. That calculation gives \(0.10\), but it evaluates a different event.
The correction is to translate the definition literally: include every possible pair satisfying both weak inequalities. This matters whenever a discrete variable has positive probability at a boundary. For continuous variables, a single boundary point often has probability zero, but that does not change the definition; the CDF still uses “less than or equal to.”
Check Your Understanding
For each problem, translate the thresholds into the event they describe before calculating.
- A discrete pair \((X,Y)\) has probabilities \(P(0,0)=0.12\), \(P(0,1)=0.18\), \(P(1,0)=0.28\), \(P(1,1)=0.22\), and \(P(2,1)=0.20\). Find \(F_{X,Y}(1,0)\). (Answer: 0.40)
- A hypothetical review process has possible pairs \((X,Y)=(1,0),(2,1),(3,0),(3,1)\) with probabilities \(0.25,0.30,0.15,0.30\), respectively. Find \(F_{X,Y}(2.5,1)\). (Answer: 0.55)
- Let \(U\sim\operatorname{Uniform}(0,1)\), \(X=U\), and \(Y=1-U\). Find \(F_{X,Y}(0.65,0.55)\). (Answer: 0.20)
- For the same \(X\) and \(Y\) as in question 3, find \(F_{X,Y}(0.35,0.55)\). (Answer: 0)
- A discrete pair has mass \(0.16\) at \((1,0)\), \(0.24\) at \((1,1)\), \(0.20\) at \((2,0)\), and \(0.40\) at \((3,1)\). Find \(F_{X,Y}(2,1)\), and distinguish it from \(P(X=2,Y=1)\). (Answer: \(F_{X,Y}(2,1)=0.60\), while \(P(X=2,Y=1)=0\))