Вы находитесь на странице: 1из 6
Tdeas OE SMT ITs / 3. LERey Poles q Eh. suite INTERPRETATION OF PROBABILITY 7.1 Probability has many meanings. Those who use the English language are accustomed to words with multiple definitions, where the meaning, depends on the context. However, it sometimes comes as a shack that words used in technical theories are not uniquely defined. How can words such as electron, mass, energy, and force have different meanings? How can probability have different interpretations? The word probability is used basically in wo different ways: (I) an intrinsic property of some system that does not depend on our knowledge of that system, and (2) a measure of belief in some statement, For years a controversy has existed among some probabilists and statisticians about the proper meaning of the word probability. At other times bitter controversy has resulted from a failure to distinguish between the two basic meanings, The argument goes back at least to von Mises [8] in 1919 and Keynes [4] in 1921 and continues to the present. The situation is illustrated by two books in elementary statistics. Savage [7] begins his hook with the view that probability is a quantitative measure of uncertainty. Noether [5] presents the degree of belief interpretation as a secondary interpretation. The different uses of the word probability lie at the heart of the Bayesian controversy, which we discuss in Lecture 31, 67.2 Scientific theories are built on undefined (or ill-defined) terms, Attempts lo define mass, force, and acceleration are unsatisfactory. Yet airplanes fly, locomotives move, and satellites orbit the earth because of theories based on these terms, Electrons have sometimes been described as particles, sometimes as waves, sometimes as both, and transistor technology advances even though the word electron is not well defined. 63 64 7.3 IDEAS OF STATISTICS An analogous situation exists wilh probability. Statistical methods and probability models have proven themselves useful, although probability itself has not been well defined. Itis ugeful to distinguish between the concepts we are trying to describe and our attempted descriptions (models). Table 7.1 emphasizes this distinction, Tia lecturee asks a class for the probability that a regular coin will land heads, the answer will probably be a resounding “one-half.” (Note that the previous sentence uses probability both asa property of the coin and ‘as a measure of degree of belief.) The answer of. one-halfis undoubtedly a conditioned response; nevertheless, it is clear that the answer is an attempt to describe a property of the coin and not a degree of belief, The properly is the long-run proportion of heads, or long-run telative frequency. To illustrate the idea, | tossed a coin 50 times, ‘The results are shown in Table 7.2, The number of the tass is shown in the first column and is denoted by n. The result of the toss is shown in the second column under H or T, The total number of heads at any given tossisshown in the third column and is denoted by n(E). Finally, the relative frequency of heads, calculated by a(A)/n, #s given in the fourth eolumn. These results are shown graphically in Figure 7.1. Note that after 50 tosses the relative frequency has settled down in the neighborhood of one-half, We would suspect that if we tossed further, the relative frequency would fluctuate ‘above and below one-half but would settle even closer to the long-run relative frequency of one-half, ‘There are several coin-tossing experiments recorded in the literature. ‘Awwell-known example is given by Kerrick [3]. While interned at Hald. Denmark, during World War Il, he performed several experiments. For example, he tossed a coin 10,000 times and: made 5000 draws of two balls from an urn. Initially in the coin-tossing experiment the relative frequency of heads fluctuated widely but, after 10,000 tosses, it settled down in the vicinity of one-half, with a value of 0,507 on the final toss, Table 7.1 [Interpretations and Models of Probability Interpretation Model Property of a system 1. Mathematical limit (long-run frequency) 2. Collectives of von Mises Measurement of degree of beliel 1. Logical relation theory ‘ 2. Subjective probability 63 INTERPRETATION OF PROBABILITY Table 7.2. Relative Frequency of Heads a Hort a() alain n fort met) n(A)én 1 H 1+ 40000 «26 12 0.4615 2 H 2 10000 270 OF 12 O.aaaa 3 H 3 10000 28 # 13° 0.4693 4 T 3 07600 29 7 18 0.4483 5 H 4 og000 30 # 14 0.4667 6 HW 5 08333 31 # 15 0.4839 7 H 6 0.8871 32H 16 0.5000 a r 6 07500 33 F 16 0.4848 9 r 6 068667 34 TF 16 © 0.4706 10 H 7 07000 35 =F 16 © 0.4571 "1 H 8 072773 36 # 17 0,4722 12 H 9 07500 37 TF 170.4595 13 7 9 06923 38 TF 17 0.4474 14 H 100071480 89 170.4389 16 r 10 06667 40 F | 17 0.4250 16 r 10 06250 410 18_ 0.4390 17 i 1005882 4«420—COF 18° 0.4286 18 c 10 0.5556 43S 18 0.4186 19 7 10 05263 44 =F 18 0.4091 20 7 40° 0.500045 19 0.4222 21 a 11 05238046 20 © 0.4348 22 7 110 05000 47 21 0.4468 23 r 11 «O4783 48 22 0.4583 24 H 12 0.5000 49 r 22 0.4490 25 z 12 04800 «500 f 23 © 0.4600 aillilin Le i l = L 1 | | Ly 5 10 15 2 2 3 08 4 45° 50 Figure 7.1 Relative Frequency of Heads.

Вам также может понравиться