Why Independence and Support Come First
Before calculating a sum distribution, determine two things: whether the summands are independent and which values each summand can take with positive probability. Independence controls which pairs of values can occur together. The supports control which sums those pairs can produce. Skipping either check can lead to a plausible calculation for the wrong model.
These checks are useful even when a problem later calls for a formula specific to a distribution. For example, recognizing two marginal distributions does not justify adding their distributions as if the variables were independent. Nor does an informal range calculation always describe the values that have positive probability. This tutorial develops the checks themselves; it does not assume a particular formula for computing a sum distribution.
For a continuous random variable, individual point probabilities are generally zero, so the same definition would not be useful. Its support is usually understood as the set of points for which every open neighborhood has positive probability, together with any boundary points approached by such neighborhoods. In particular, a continuous variable can have a support containing an endpoint even though the probability of equaling that endpoint is zero.
Check Independence Before Combining Supports
Two random variables \(X\) and \(Y\) are independent when their joint probabilities factor into their marginal probabilities. In particular, for discrete variables, independence requires \(P(X=x,Y=y)=P(X=x)P(Y=y)\) for every pair \(x,y\). For general random variables, independence means that this factorization holds for all events determined by \(X\) and \(Y\), not merely for one selected pair.
The key practical consequence is that, for independent discrete variables, every pair of values with positive marginal probabilities has positive joint probability. If \(P(X=x)>0\) and \(P(Y=y)>0\), then \(P(X=x,Y=y)=P(X=x)P(Y=y)>0\). If the variables are dependent, some such pairs may be ruled out, and adding every value in one support to every value in the other can overstate the sum’s support.
Identify the variables and state whether independence is given or can be established from their joint behavior. Different marginal distributions do not establish independence.
For discrete variables, include exactly the values with positive probability. Keep zero-probability values out of the positive-probability support.
When the variables are independent and discrete, add every support value of one variable to every support value of the other. Remove duplicates.
Distinguish sums with positive probability from values that might occur only on a probability-zero outcome. For continuous variables, check the support as a range with its boundary behavior.
The Discrete Sum-Support Result
Here is why the independent case holds. For any \(x\in\mathcal S_X\) and \(y\in\mathcal S_Y\), independence gives \(P(X=x,Y=y)=P(X=x)P(Y=y)>0\). Since that joint event implies \(X+Y=x+y\), the sum \(x+y\) has positive probability. Conversely, suppose \(P(X+Y=s)>0\). For discrete variables, the event \(\{X+Y=s\}\) is the union of the countably many pair events \(\{X=x,Y=y\}\) for which \(x+y=s\). If every one of those pair events had probability zero, their countable union would also have probability zero. Therefore at least one pair has positive joint probability. Independence then implies \(P(X=x)>0\) and \(P(Y=y)>0\), so \(s\) is a sum of support values. This argument concerns positive probability; it does not claim that a value outside the support cannot occur on a null event.
Worked Example: Independent Discrete Variables with a Repeated Sum
Let \(X\) and \(Y\) be independent. Suppose \(P(X=0)=0.2\), \(P(X=2)=0.5\), \(P(X=5)=0.3\), and \(P(Y=1)=0.4\), \(P(Y=3)=0.6\). Find the positive-probability support of \(X+Y\), and calculate \(P(X+Y=3)\). The stated probabilities are the complete distributions of the two variables.
Method. Since independence is given, form all pairwise sums of support values. For the probability of a particular sum, add the probabilities of the disjoint pairs that produce it, using independence for each pair.
The supports are \(\mathcal S_X=\{0,2,5\}\) and \(\mathcal S_Y=\{1,3\}\). The six pairs give \(0+1=1\), \(0+3=3\), \(2+1=3\), \(2+3=5\), \(5+1=6\), and \(5+3=8\). Thus \(\mathcal S_{X+Y}=\{1,3,5,6,8\}\). The value 3 appears twice, so both pairs must be included in its probability: \(P(X+Y=3)=P(X=0,Y=3)+P(X=2,Y=1)\). By independence, this equals \((0.2)(0.6)+(0.5)(0.4)=0.12+0.20=0.32\).
Check. Each of the five listed sums has at least one pair with positive probability. The six pair probabilities total \((0.2+0.5+0.3)(0.4+0.6)=1\), and the repeated sum 3 correctly combines two disjoint pairs. The answer is \(0.32\).
What Changes When the Variables Are Dependent?
If variables are dependent, their marginal supports alone do not say which pairs can occur together. The relevant pairs are those with positive joint probability. In particular, a pair can consist of two values that each have positive marginal probability and still have joint probability zero.
Worked Example: Marginal Supports Do Not Guarantee Independent Pairs
A hypothetical two-state mechanism produces \((X,Y)=(0,1)\) with probability \(0.5\) and \((X,Y)=(1,0)\) with probability \(0.5\). Find the supports of \(X\), \(Y\), and \(X+Y\). Explain why the independent-support calculation would be wrong.
From the two outcomes, \(P(X=0)=P(X=1)=0.5\), so \(\mathcal S_X=\{0,1\}\). Likewise, \(P(Y=0)=P(Y=1)=0.5\), so \(\mathcal S_Y=\{0,1\}\). But the joint pairs are only \((0,1)\) and \((1,0)\). Both produce a sum of 1, so \(P(X+Y=1)=1\) and \(\mathcal S_{X+Y}=\{1\}\).
The variables are not independent: \(P(X=0,Y=0)=0\), whereas \(P(X=0)P(Y=0)=(0.5)(0.5)=0.25\). If we incorrectly treated them as independent, the pairwise sums of the marginal supports would be \(\{0,1,2\}\), even though neither 0 nor 2 can occur in this model. The dependence restriction is essential.
Check. Each possible joint outcome has probability \(0.5\), and the sum equals 1 in both outcomes. Therefore the stated sum support and probability agree with the complete joint model.
Continuous Supports and Their Endpoints
For continuous variables, think in terms of the ranges where probability is concentrated, rather than asking whether each single value has positive probability. For independent variables with interval supports, adding the interval endpoints identifies the interval containing the sum’s support. Endpoints may belong to that support even though the probability of hitting either endpoint exactly is zero.
Worked Example: Adding Two Uniform Supports
Let \(X\sim\operatorname{Uniform}(1,3)\) and \(Y\sim\operatorname{Uniform}(2,5)\), independently. Identify the support of \(X+Y\), and state whether the sum can equal either endpoint with positive probability.
Model and assumptions. \(X\) has a continuous uniform distribution on \([1,3]\), \(Y\) has a continuous uniform distribution on \([2,5]\), and they are independent. The sum is bounded below by \(1+2=3\) and above by \(3+5=8\).
Every interior value \(s\) with \(3<s<8\) can be represented as \(x+y=s\) with \(1<x<3\) and \(2<y<5\). To see this, choose \(x\) in the interval \((\max(1,s-5),\min(3,s-2))\). This interval is nonempty for \(3<s<8\); then \(y=s-x\) lies strictly between 2 and 5. A small rectangle around that pair has positive probability under the independent uniform model, so neighborhoods of \(s\) have positive probability.
The sum’s support is therefore \([3,8]\). However, \(X+Y=3\) requires \(X=1\) and \(Y=2\), while \(X+Y=8\) requires \(X=3\) and \(Y=5\). Each endpoint event has probability zero for continuous uniform variables. The endpoints are in the support because values arbitrarily near them have positive probability, not because the endpoints themselves have positive point probability.
Check. The sum cannot be below 3 or above 8, and interior values have positive probability in every neighborhood. Thus \([3,8]\) is the support, with zero probability at either endpoint.
Common Mistake: Treating Every Pair as Possible
A frequent error is to take the pairwise sums of marginal supports without first checking independence. In the dependent two-state example, the tempting calculation is \(\{0,1\}+\{0,1\}=\{0,1,2\}\). That is the wrong calculation for the given model: it assumes all four pairs are possible. The correction is to use the joint distribution; only \((0,1)\) and \((1,0)\) have positive probability, so the sum support is \(\{1\}\).
Another subtle error is to treat “not in the positive-probability support” as equivalent to “cannot occur on any outcome.” A random variable may take an unusual value on a null event. Such a value has probability zero and is not part of its positive-probability support, but it may still be a pointwise value in a chosen sample-space description. For probability calculations, state conclusions in terms of positive probability or probability zero unless the model explicitly rules out the value on every outcome.
These checks are aligned with the SOA Exam P (Probability) syllabus because they determine whether a standard independent-sum approach fits the stated model before any further calculation begins. A concise solution should identify independence, list the supports, form the allowed sums, and state whether its conclusion concerns positive probability or pointwise possibility.
Check Your Understanding
For each problem, check the assumptions before forming or interpreting the sum support.
- \(X\) and \(Y\) are independent, with \(\mathcal S_X=\{-1,2\}\) and \(\mathcal S_Y=\{0,3\}\). List the positive-probability support of \(X+Y\), including any repeated sum only once. (Answer: \(\{-1,2,5\}\))
- Independent variables have \(P(X=0)=0.25\), \(P(X=4)=0.75\), \(P(Y=-1)=0.6\), and \(P(Y=2)=0.4\). Find the positive-probability support of \(X+Y\) and \(P(X+Y=3)\). (Answer: support \(\{-1,2,3,6\}\); probability \(0.4500\))
- A random pair is \((X,Y)=(2,0)\) with probability \(0.3\) and \((X,Y)=(0,2)\) with probability \(0.7\). Find the marginal supports, decide whether independence holds, and find the sum support. (Answer: \(\mathcal S_X=\mathcal S_Y=\{0,2\}\); dependent; sum support \(\{2\}\))
- Independent variables are uniform on \([0,2]\) and \([1,4]\), respectively. Give the support of their sum and the probability that the sum equals 1 exactly. (Answer: support \([1,6]\); probability \(0\))
- On a sample space, \(\omega_0\) has probability 1 and \(\omega_1\) has probability 0. Let \(X(\omega_0)=0\), \(X(\omega_1)=7\), and let \(Y=2\) on both outcomes. Are \(X\) and \(Y\) independent? Give their positive-probability supports, the positive-probability support of the sum, and the additional pointwise sum on the null outcome. (Answer: yes; \(\mathcal S_X=\{0\}\), \(\mathcal S_Y=\{2\}\), positive-probability sum support \(\{2\}\), and pointwise sum \(9\) on \(\omega_1\))