PROBABILITY THEORY:SAMPLE SPACE

SAMPLE SPACE

Suppose the demand for a product over a period of time, say a month, is of interest. From a realistic point of view, demand is not generally constant but exhibits the type of variation alluded to in the introduction. Suppose an experiment that will result in observing the demand for the product during the month is run. Whereas the outcome of the experiment cannot be predicted exactly, each possible outcome can be described. The demand during the period can be any one of the values 0, 1, 2, . . . , that is, the entire set of nonnegative integers. The set of all possible outcomes of the experiment is called the sample space and will be denoted by n. Each outcome in the sample space is called a point and will be de- noted by w. Actually, in the experiment just described, the possible demands may be bounded from above by N, where N would represent the size of the population that has any use for the product. Hence, the sample space would then consist of the set of the integers 0, 1, 2, . . . , N. Strictly speaking, the sample space is much more complex than just described. In fact, it may be extremely difficult to characterize precisely. Associated with this experiment are such factors as the dates and times that the demands occur, the prevailing weather, the disposition of the personnel meeting the demand, and so on. Many more factors could be listed, most of which are irrelevant. Fortunately, as noted in the next section, it is not necessary to describe completely the sample space, but only to record those factors that appear to be necessary for the purpose of the experiment.

Another experiment may be concerned with the time until the first customer arrives at a store. Since the first customer may arrive at any time until the store closes (assuming an 8-hour day), for the purpose of this experiment, the sample space can be considered to be all

points on the real line between zero and 8 hours. Thus, n consists of all points w such that 0 ::: w ::: 8.†

Now consider a third example. Suppose that a modification of the first experiment is made by observing the demands during the first 2 months. The sample space n consists of all points (x1,x2), where x1 represents the demand during the first month, x1 = 0, 1, 2,

. . . , and x2 represents the demand during the second month, x2 = 0, 1, 2, . . . . Thus, n consists of the set of all possible points w, where w represents a pair of nonnegative in- teger values (x1,x2). The point w = (3,6) represents a possible outcome of the experiment where the demand in the first month is 3 units and the demand in the second month is 6 units. In a similar manner, the experiment can be extended to observing the demands during the first n months. In this situation n consists of all possible points w = (x1, x2, . . . , xn), where xi represents the demand during the ith month.

The experiment that is concerned with the time until the first arrival appears can also be modified. Suppose an experiment that measures the times of the arrival of the first cus- tomer on each of 2 days is performed. The set of all possible outcomes of the experiment n consists of all points (x1,x2), 0 ::: x1, x2 ::: 8, where x1 represents the time the first cus- tomer arrives on the first day, and x2 represents the time the first customer arrives on the second day. Thus, n consists of the set of all possible points w, where w represents a point in two space lying in the square shown in Fig. 24.1.

This experiment can also be extended to observing the times of the arrival of the first customer on each of n days. The sample space n consists of all points w = (x1, x2, . . . , xn), such that 0 ::: xi ::: 8 (i = 1, 2, . . . , n), where xi represents the time the first customer ar- rives on the ith day.

An event is defined as a set of outcomes of the experiment. Thus, there are many events that can be of interest. For example, in the experiment concerned with observing the demand for a product in a given month, the set {w = 0, w = 1, w = 2, . . . , w = 10} is the event that the demand for the product does not exceed 10 units. Similarly, the set {w = 0} denotes the event of no demand for the product during the month. In the exper- iment which measures the times of the arrival of the first customer on each of 2 days, the set {w = (x1, x2); x1 < 1, x2 < 1} is the event that the first arrival on each day occurs be- fore the first hour. It is evident that any subset of the sample space, e.g., any point, col- lection of points, or the entire sample space, is an event.

Events may be combined, thereby resulting in the formation of new events. For any two events E1 and E2, the new event E1 U E2, referred to as the union of E1 and E2, is †It is assumed that at least one customer arrives each day.

defined to contain all points in the sample space that are in either E1 or E2, or in both E1 and E2. Thus, the event E1 U E2 will occur if either E1 or E2 occurs. For example, in the demand experiment, let E1 be the event of a demand in a single month of zero or 1 unit, and let E2 be the event of a demand in a single month of 1 or 2 units. The event E1 U E2 is just {w = 0, w = 1, w = 2}, which is just the event of a demand of 0, 1, or 2 units.

The intersection of two events E1 and E2 is denoted by E1 n E2 (or equivalently by E1E2). This new event E1 n E2 is defined to contain all points in the sample space that are in both E1 and E2. Thus, the event E1 n E2 will occur only if both E1 and E2 occur. In the aforementioned example, the event E1 n E2 is {w = 1}, which is just the event of a demand of 1 unit.

Finally, the events E1 and E2 are said to be mutually exclusive (or disjoint) if their in- tersection does not contain any points. In the example, E1 and E2 are not disjoint. How- ever, if the event E3 is defined to be the event of a demand of 2 or 3 units, then E1 n E3 is disjoint. Events that do not contain any points, and therefore cannot occur, are called null events. (Or course, all these definitions can be extended to any finite number of events.)