ESTIMATION A. Armed w i t h t h e c e n t r a l l i m i t theorem and t h e normal d i s t r i b u t i o n , one can o b t a i n t h e p r o b a b i l i t y t h a t one's ESTIMATE o f a p o p u l a t i o n parameter i s i n error. - 1. I f sample means ( i .e., Ys) are n o r m a l l y d i s t r i b u t e d about a p o p u l a t i o n parameter, p, then we know t h a t 95% o f a l l Vs t a k e values on t h e interval, p f 1.96 u y / h . h 2. What we have i n p r a c t i c e , however, a r e n o t p and uy, b u t 7 and uy which a r e used as p o i n t estimates o f p and uy. 3. I n e s t i m a t i n g these parameters we want estimates t h a t a r e b o t h UNBIASED (centered around t h e parameter) and EFFICIENT (having a small degree o f sampling e r r o r r e l a t i v e t o o t h e r e s t i m a t o r s ) . For example, Y+ 10 i s a biased e s t i m a t o r o f p (although i t i s an unbiased e s t i m a t o r o f p + 10). Also Y1 i s n o t as e f f i c i e n t as 7 in 2 whereas t h e variance o f e s t i m a t i n g p, because t h e variance o f Y1 i s uy, 4. As you might expect, Y i s an e f f i c i e n t , unbiased e s t i m a t o r o f p. 2 i s NOT an UNbiased e s t i m a t o r o f uy: 2 However, t h e sample variance, sy, To demonstrate t h i s , i t must f i r s t be acknowledged t h a t l / n Z (Yi - p)2 2 2 i s t h e average squared Recall t h a t uy i s an unbiased e s t i m a t o r o f uy. d e v i a t i o n from t h e p o p u l a t i o n mean. v a r i a b l e s (e.g., L i k e averages f o r o t h e r random - Y f o r Y), a sample average o f squared d e v i a t i o n s from p i s an unbiased e s t i m a t o r o f t h e p o p u l a t i o n average o f such squared deviations. By showing t h a t t h i s unbiased e s t i m a t o r (namely, 65 2 i t can be demonstrated l / n 1 (Yi - p ) 2 ) i s c o n s i s t e n t l y l a r g e r t h a n sy, 2 t h a t sy2 i s biased as an e s t i m a t o r o f ay. Thus, t h e t h e s i s i s t h a t f o r any sample We begin by n o t i n g t h a t Since n(Y - fi12 2 0 , t h e p r o o f i s complete. Thus " s y M c o n s i s t e n t l y UNDERestimates ay. We can a d j u s t f o r t h i s b i a s by changing t h e denominator o f s t from n t o n-1. "2 uy = T h i s i s why l(Yi - Y) 2 and n o t st i s used as t h e POINT ESTIMATE o f a t . n - 1 (REMEMBER: The " h a t " n o t a t i o n i n d i c a t e s t h a t 5. WARNING: :t i s an e s t i m a t o r o f a t . ) When e s t i m a t i n g t h e v a r i a n c e o f X, d i v i d e by "n-1"; when e s t i m a t i n g t h e v a r i a n c e o f T(, d i v i d e by "n". Regarding t h e l a t t e r warning, r e c a l l t h a t according t o t h e c e n t r a l l i m i t theorem, 2 2 "X - - "X - n NOT 2 ax !!! n-1 B. We have a l r e a d y t a l k e d about INTERVAL ESTIMATION i n which a CONFIDENCE INTERVAL i s estimated. But i n t e r v a l s are o n l y meaningful along some 66 continuum o f values. Thus i t seems p e c u l i a r t o speak o f c o n s t r u c t i n g an i n t e r v a l around a "mean" f o r NOMINAL data. up t h e numbers and d i v i d e by "n." You CANNOT, o f course, j u s t add (Reminiscent o f o u r e a r l i e r d i s c u s s i o n on LEVELS OF MEASUREMENT, t h i s would be "1 ike adding apples and oranges. " ) Our s o l u t i o n w i l l be t o consider & cateqorv s e ~ a r a t e l y . Taking a v a r i a b l e such as r e l i g i o u s a f f i l i a t i o n , f o r example, a new v a r i a b l e ( c a l l i t "D") c o u l d be c o n s t r u c t e d such t h a t D= NOTE: { 0 1 when n o t C a t h o l i c when C a t h o l i c V a r i a b l e s l i k e t h i s (namely, t h a t t a k e t h e value, one, when s u b j e c t s have an a t t r i b u t e and zero otherwise) are c a l l e d dummv variables. C. ESTIMATING PROPORTIONS: 1. Applying t h e formula f o r t h e mean t o a dummy v a r i a b l e always produces a number w i t h a value somewhere between zero and one. For example, i f t h e average o f t h e C a t h o l i c a f f i l i a t i o n v a r i a b l e were .6 , 67 t h i s would mean t h a t 60% o f t h e C a t h o l i c s i n your sample a r e C a t h o l i c . This f o l l o w s because t h e mean o f t h e Ds i s n = - n 2 Di i=l n ' n . But s i n c e D D = = = 0 when D # C a t h o l i c and 1 when D = C a t h o l i c then p r o p o r t i o n of C a t h o l i c s i n t h e sample. h 2. Now l e t ' s consider t h e variance o f t h i s s t a t i s t i c , n. You w i l l r e c a l l t h a t t h e peg-board i l l u s t r a t i o n (used when we i n t r o d u c e d t h e concept o f n o r m a l i t y ) was a demonstration o f t h e sampling d i s t r i b u t i o n o f a proportion. Applying t h e lesson from t h a t i l l u s t r a t i o n t o t h i s dummy v a r i a b l e , we know t h a t i f one were t o draw a l o t o f random samples o f "nu r e s i d e n t s from t h e c i t y o f Antwerp, one would expect t h a t t h e i r r e s p e c t i v e C a t h o l i c - p r o p o r t i o n s would be d i s t r i b u t e d n o r m a l l y ( f o r l a r g e n) around t h e t r u e p r o p o r t i o n o f C a t h o l i c s i n t h e p o p u l a t i o n . To be able t o make p r o b a b i l i s t i c inferences about t h e p r o p o r t i o n o f C a t h o l i c s t h a t we f i n d i n our sample, we must be more s p e c i f i c about t h e exact n a t u r e o f t h e sampling d i s t r i b u t i o n o f these sample p r o p o r t i o n s : a. Applying t h e formula f o r t h e variance t o t h e dummy v a r i a b l e , D: N o t i n g t h a t when n = 0 or 1 , D = , D~ =Di A And since D = - n ' F i n a l l y , n o t e t h a t when t h e sample s i z e i s l a r g e , t h e r a t i o n n - 1 , can ( f o r a l l p r a c t i c a l purposes) be considered equal t o 1. A Thus A 2 uD = nA ( 1 - n ) f o r l a r g e samples. b. But t h i s i s t h e large-sample estimate o f t h e variance o f a dummy The c e n t r a l l i m i t theorem i s a p p l i e d variable, not o f a proportion. A t o n n e a r l y i d e n t i c a l l y as i t was a p p l i e d t o X. That i s , t h e variance o f a p r o p o r t i o n equals t h e variance o f i t s associated dummy I n o t h e r words, t h e variance o f v a r i a b l e d i v i d e d by t h e sample size. a p r o p o r t i o n i s estimated as f o l l o w s : A Accordingly, a - N( n, n(1 - n) n ) i s what i s c a l l e d t h e "normal approximation o f t h e binomial d i s t r i b u t i o n . " BINOMIAL DISTRIBUTION s h o r t l y . 69 We s h a l l r e t u r n t o t h e c. One more comment about e s t i m a t i n g p r o p o r t i o n s : 1 The f o l l o w i n g i s a t h r e e - p a r t " r u l e o f thumb" f o r d e c i d i n g when your sample i s " l a r g e enough" f o r you t o assume t h a t "i" and " 1 - a" A 1) Take whichever i s SMALLER o f and 2) m u l t i p l y i t by "n". A 3) I f t h i s number i s g r e a t e r than f i v e , you may assume t h a t a i s normally d i s t r i b u t e d . D. Sample s i z e r e q u i r e d i n e s t i m a t i n g p r o p o r t i o n s 1. CBS/N.Y. Times telephone p o l l e r s c l a i m t o have estimates o f p u b l i c o p i n i o n t h a t are i n e r r o r by o n l y How - + 3% p o i n t s . Consider t h e question, a samole would you need t o estimate t h e p r o p o r t i o n o f Americans t h a t would answer "Yes" t o t h e question, "Do you agree w i t h t h e P r e s i d e n t ' s economic p o l i c i e s ? " 2. Put d i f f e r e n t l y , you want your estimate o f t h e p r o p o r t i o n t o have a PRECISION (i.e., + 3% ! t h e s u b s t a n t i v e e r r o r one i s w i l l i n g t o t o l e r a t e ) o f NOTE t h a t saying t h a t you want y o u r e s t i m a t e t o be w i t h i n 3% o f t h e t r u e amount SAYS NOTHING about t h e s i g n i f i c a n c e l e v e l ( o r a ) . A a. We know t h a t a 100(1-a)% confidence i n t e r v a l around a would be T h i s r u l e o f thumb i s from Alan A g r e s t i and Barbara F i n l a y ( S t a t i s t i c a l Methods f o r t h e S o c i a l Sciences, 2nd e d i t i o n . San Francisco, CA: D e l l e n , 1986, p. 142.). However, see a l s o Alan A g r e s t i and Brent A. C o u l l (1998. "Approximate i s B e t t e r than 'Exact' f o r I n t e r v a l E s t i m a t i o n o f Binomial Proportions." Amercian S t a t i s t i c i a n 52:119-126). b. It i s t h e l a s t p a r t o f t h i s expression t h a t i s t h e p r e c i s i o n t h a t t h e CBS/New York Times p o l l e r s set a t 3 percent. I n accordance w i t h my e a r l i e r d i s c u s s i o n o f t h e B i g P i c t u r e ( L e c t u r e Notes, p. 36), t h e Greek c a p i t a l l e t t e r d e l t a , A, r e f e r s t o p r e c i s i o n . Note t h a t p r e c i s i o n i s o n e - h a l f t h e w i d t h o f a confidence i n t e r v a l . PRECISION = A = That i s , ZQ/2 REMEMBER: The l a r g e r your p r e c i s i o n i s , t h e less p r e c i s e (i.e., the more open t o d e v i a t i o n from t h e t r u e value) y o u r e s t i m a t e w i l l be! c. A f t e r s e l e c t i n g a p r e c i s i o n o f l e t ' s say, Q = .05 , A = 3% and a s i g n i f i c a n c e l e v e l o f , t h i s formula r e q u i r e s t h a t we a l s o have a A v a l u e f o r n t o determine how l a r g e a sample i s needed. A The most c o n s e r v a t i v e choice when choosing a v a l u e f o r n i s a v a l u e A A t h a t y i e l d s t h e l a r g e s t value o f n ( 1 - n ) . T h i s a l l o w s one t o e s t i m a t e t h e l a r g e s t p o s s i b l e confidence i n t e r v a l (and, e q u i v a l e n t l y , t h e l a r g e s t p o s s i b l e p r e c i s i o n ) t h a t c o u l d r e s u l t once one's d a t a a r e collected. is A A n = h n(1 - n) I t i s easy t o v e r i f y t h a t t h i s most c o n s e r v a t i v e c h o i c e .50 ! One need merely consider, f o r example, t h e values o f A as n ranges from 0 t o 1: Thus, we f i n d t h a t r e w r i t t e n as Thus, n = n 1068 .25 = .50(1 - .50) .03 = 1.96 [ 1.96 o 3 ]2 = 1067.11 , , whichcanbe o r 1068 . i s t h e s i z e o f CBS/NYT samples! THREE COMMENTS: 1) As always, round UD t o whole numbers i n e s t i m a t i n g sample s i z e s f o r a given level o f precision. (I.e., By rounding down, a l a r g e r ( i .e., less precise) precision i s a t t a i n e d t h a n t h e one t h a t i s sought. cases, round no h a l f people, please.) (A REMINDER: I n a l l o t h e r t o t h e nearest decimal v a l u e as p e r p. 3 1 o f these L e c t u r e Notes.) 2) N o t i c e t h a t t h i s number ( i e . , LARGE THE POPULATION!!! -3% ( a t a = .05 ) , 1068) i s t h e same NO MATTER That i s , t o o b t a i n a p r e c i s i o n o f HOW A = you would need t o sample 1068 people no m a t t e r i f they were drawn from a town o f 20,000 people o r from a l l people i n t h e world. 72 a) O f course, t h e reason why few surveys o f t h e w o r l d are done i s due t o t h e d i f f i c u l t y i n o b t a i n i n g a l i s t o f a l l humans from which such a sample c o u l d be drawn. b) Although a l a r g e r sample i s n o t r e q u i r e d f o r l a r g e p o p u l a t i o n s , a s m a l l e r sample i s allowed when one's sample s i z e i s more than one t e n t h t h e s i z e o f one's p o p u l a t i o n . (For more d e t a i l s , see " f i n i t e p o p u l a t i o n c o r r e c t i o n " i n A g r e s t i and F i n l a y [1986, P 871.) 2 2 = .25): 3) A few words on e s t i m a t i n g oD (or, when t o assume oD a) I f you have c o l l e c t e d your data, you w i l l be a b l e t o c a l c u l a t e A n. A 2 I n such cases use oD reference: A = nA ( l - n ) You w i l l always use A 2 . (As a p o i n t o f f u t u r e A A oD = n ( l - n ) when f i n d i n g confidence i n t e r v a l s . ) b) I f you are t e s t i n g a hypothesis, you w i l l have decided on a "hypothesized value" o f n ( c a l l i t n o ) . = n o ( l - no) . NOTE: We s h a l l t a l k a l o t more about no when we discuss hypothesis t e s t i n g . reference: "2 I n such cases, use oD (ALSO, as a p o i n t o f f u t u r e "2 You w i l l always use oD = no(l -no) in c a l c u l a t i n g p - v a l ues and Type I 1 e r r o r s . ) A c ) When you have n e i t h e r n n o r no (as i s o f t e n t h e case when "2 choosing n), use t h e most c o n s e r v a t i v e estimate o f oD, namely A d) When you have both r and ro (as o f t e n happens when you are w r i t i n g up your f i n d i n g s ) , use t h e more c o n s e r v a t i v e e s t i m a t e "2 o f aD, namely whichever i s t h e l a r a e r o f 73 A 2 A A aD = r ( l - n ) or aA 2 = 1 n o . (WARNING: As mentioned i n "a" and "b" above, t h e r e a r e exceptions t o t h i s r u l e when f i n d i n g confidence i n t e r v a l s , p-values, and Type I 1 e r r o r s . ) E. You may r e c a l l from t h e second s e c t i o n o f these l e c t u r e notes t h a t n e i t h e r t h e C e n t r a l L i m i t Theorem n o r t h e B i g P i c t u r e h o l d s f o r estimates based on small samples. The binomial d i s t r i b u t i o n i s a case i n p o i n t . L e t us c o n s i d e r what t o do i f we f i n d t h a t t h e sample i s NOT l a r g e enough t o assume t h a t A n - N( n, n(1 - n) n 1 . That i s , what should be done i f we f i n d t h a t n * A n ( 5 or n * A (1 - n) s 5 ? To answer t h i s , we need t o t h i n k a l i t t l e more about t h e d i s t r i b u t i o n o f D = { 01 if C n oatt hCoal itch o l i c N o t i c e t h a t t h i s random v a r i a b l e a c t s much l i k e f l i p p i n g a c o i n . t h a t you have a p e r f e c t l y symmetric c o i n ( i e . , o f g e t t i n g a "heads" o r a " t a i l s , " "heads" o f n = OR Assume a c o i n w i t h equal chances with a probability o f getting With a sample o f s i z e n = 2 c o i n tosses, n o t e t h a t .5 ) . t h e d i s t r i b u t i o n a l r e a d y begins t o t a k e on a somewhat normal shape w i t h A Pr( n = 0 ) = .25 , no heads A Pr( n = .5 ) = .5 , A and one head Pr( n = 1 ) = .25 . two heads T h i s p r o b a b i l i t y d i s t r i b u t i o n can be expressed as t h a t o f a BINOMIAL RANDOM A VARIABLE, Y , by m u l t i p l y i n g each n value by t h e sample s i z e . The r e s u l t i n g d i s t r i b u t i o n i s then Pr( Y = 0 ) = .25 , P r ( Y = 1 ) = .50 , and P r ( Y = 2 ) = .25 . NOTE: The BINOMIAL RANDOM VARIABLE, Y, i s t h e sum o f t h e values taken by 74 each o f n independently sampled observations on a dummy v a r i a b l e . The sampling d i s t r i b u t i o n o f a binomial random v a r i a b l e i s c a l l e d t h e BINOMIAL PROBABILITY DISTRIBUTION. The general form o f t h i s b i n o m i a l d i s t r i b u t i o n i s as f o l l o w s : pr( Y = Y ) No. = k] ny ( 1 - n)n-y , where k] i n! y!(n-y)! P l a c i n g an exclamation p o i n t a f t e r a number does n o t i n d i c a t e t h e s t a t i s t i c i a n ' s excitement about t h a t p a r t i c u l a r number. When an exclamation p o i n t i s placed a f t e r an i n t e g e r , t h i s r e f e r s t o a " f a c t o r i a l . " I t means t h a t you m u l t i p l y t h a t number by a l l i n t e g e r s between i t and zero. For example, " f i v e f a c t o r i a l " ( 5 ! ) i s c a l c u l a t e d as Moreover, by d e f i n i t i o n , O! = 1 5*4*3*2*1 = 120 . . T h i s seems a l o t o f t r o u b l e u n t i l you c o n s i d e r Pr( Y < 3 ) when n = .8 and n = 32 ! L e t ' s see how we would c a l c u l a t e t h i s : A f t e r c a l c u l a t i n g these numbers, CONCLUSION: Pr( Y < 3 ) equals t h e i r sum. When your sample s i z e i s t o o small f o r you t o assume t h a t t h e A sampling d i s t r i b u t i o n o f n i s normally d i s t r i b u t e d , you should use t h e binomial d i s t r i b u t i o n i n s t e a d o f t h e standard normal d i s t r i b u t i o n t o 75 determine i f y o u r r e s u l t s are s i g n i f i c a n t l y d i f f e r e n t from what you would assume would happen by chance alone. Moreover, note t h a t although t h e binomial d i s t r i b u t i o n may be computationally i m p r a c t i c a l f o r very l a r g e samples, i t ALWAYS APPLIES f o r p r o p o r t i o n s . instance, J r * n = .8*32 = 25.6 > 5 p I n t h i s l a s t example, f o r ( 1 - r ) * n = .2*32 = 6.5 > 5 , a l l o w i n g one t o use e i t h e r t h e binomial d i s t r i b u t i o n o r i t s normal approximation i n drawing i n f e r e n c e s about p o p u l a t i o n p r o p o r t i o n s . F. t-DISTRIBUTION: I f you a r e making an i n t e r v a l estimate o f t h e mean o f an i n t e r v a l - o r r a t i o - l e v e l v a r i a b l e and i f y o u r sample s i z e i s small (say, l e s s than about n = 30 ) , you w i l l need t o use t h e t - d i s t r i b u t i o n (see Table B) i n s t e a d o f t h e standard normal d i s t r i b u t i o n . 1. A l l t h a t i s being s a i d here i s t h a t ( f o r t h e purposes hand-calculated problems d u r i n g t h i s course) when n > 30, you should use y o u r standard normal t a b l e i n e s t i m a t i n g p r o b a b i l i t i e s associated w i t h sample means. When n ( 30, use t h e t - d i s t r i b u t i o n t a b l e . ( S t a t i s t i c a l software l i k e SPSS w i l l always use t h e t - d i s t r i b u t i o n i n t e r n a l l y . ) 2. There i s an important r e s t r i c t i o n when i t comes t o u s i n g t h e t d i s t r i b u t i o n , however. I f i n one's p o p u l a t i o n t h e d i s t r i b u t i o n o f a v a r i a b l e i s n o t normal, i t i s l e g i t i m a t e (given t h e c e n t r a l l i m i t theorem) t o assume t h a t t h e sampling d i s t r i b u t i o n o f one's s t a t i s t i c i s normal. I n c o n t r a s t , when your sample i s small i t i s NOT l e g i t i m a t e t o assume t h a t t h e sampling d i s t r i b u t i o n o f y o u r s t a t i s t i c i s t h a t o f a t d i s t r i b u t i o n , unless you have reason t o b e l i e v e t h a t y o u r s t a t i s t i c ' s underlying variable fi normally d i s t r i b u t e d i n the population. 3. N o t i c e t h a t t h e t a b l e f o r t h e t - d i s t r i b u t i o n i s organized s i m i l a r l y t o t h e chi-square t a b l e , w i t h degrees o f freedom down t h e l e f t column and 76 Like the chi-square table, the p r o b a b i l i t i e s a t t h e t o p o f t h e columns. t - t a b l e presents values on a s t a t i s t i c (here, t - s c o r e s i n s t e a d o f c h i square scores) f o r a l a r g e number o f d i s t r i b u t i o n s - e a c h distribution d e f i n e d according t o t h e s t a t i s t i c ' s degrees o f freedom. For example, when e s t i m a t i n g a confidence i n t e r v a l f o r a mean, t h e t - d i s t r i b u t i o n w i t h n - 1 degrees o f freedom should be used. (NOTE: The shape o f t h e t - d i s t r i b u t i o n g e t s " f l a t t e r , " t h e fewer i t s degrees o f freedom.) 4. AN EXAMPLE: Imagine t h a t a government o f f i c i a l wishes t o a p p r o p r i a t e As p a r t o f h i s planning, he wishes t o e s t i m a t e funds f o r f l o o d r e l i e f . ( a t t h e .10 l e v e l o f s i g n i f i c a n c e ) how many o f t h e people who r e s i d e on t h e M i s s i s s i p p i f l o o d p l a i n move o f f o f t h e f l o o d p l a i n w i t h i n a y e a r o f the r i v e r ' s flooding. He o n l y has d a t a f o r t h e l a s t 30 years and d u r i n g these years o n l y s i x f l o o d s have occurred. He assumes h i s s i x measures o f moved r e s i d e n t s t o have been drawn from a n o r m a l l y d i s t r i b u t e d p o p u l a t i o n o f resident-moves d u r i n g p a s t f l o o d s . Thus, h i s goal i s t o f i n d an i n t e r v a l estimate o f t h e number o f people who move a f t e r Mississippi floods. People who moved: H i s d a t a are 21, 31, 19, 38, 60, 47. C l e a r l y , t h i s sample i s smaller than n = 30 . Note t h a t i f we use t h e normal d i s t r i b u t i o n , t h e 90% confidence i n t e r v a l i s - * za/2 , which s i n c e X = A "2 = 248 [ox 36 and ox y i e l d s a 90% confidence i n t e r v a l o f 36 between 25.4 and 46.6 . + 1.645 * 15'75 fi = 15.751 o r an i n t e r v a l However, s i n c e n = 6 < 30 expression. , we must use t a/2 R e f e r r i n g t o Table B we f i n d t h a t and n o t Z a/2 i n this t5,.05 = 2.015 . Thus o u r 90% confidence i n t e r v a l becomes 36 t 2.015 15'75 fi o r an i n t e r v a l between 23.0 and 49.0 . NOTE how i n comparison t o t h e z-score, t h e t - s c o r e leads you t o make a more c o n s e r v a t i v e ( i .e., a 'wider') i n t e r v a l estimate. 5. Now l e t ' s imagine t h a t you c o u l d o b t a i n d a t a on an a d d i t i o n a l f o u r t e e n M i s s i s s i p p i f l o o d s i f you were t o go t o t h e c o n s i d e r a b l e e x t r a t i m e and expense o f o b t a i n i n g d a t a from l o c a l newspapers on f l o o d s t h a t t o o k p l a c e d u r i n g t h e previous century. Making use o f t h e d a t a from t h e s i x cases t h a t you do have, you assume t h a t t h e p o p u l a t i o n mean i s i n f a c t 36 f l o o d v i c t i m s , and t h a t t h e p o p u l a t i o n v a r i a n c e equals 248 (squared victims). Here's a q u e s t i o n f o r you: "How f a r from ( t h e assumed p o p u l a t i o n mean o f ) 36 would t h e mean from a random sample o f 20 f l o o d s have t o be f o r i t t o have a p r o b a b i l i t y o f l e s s than .05 o f having been sampled from t h i s assumed p o p u l a t i o n o f M i s s i s s i p p i R i v e r f l o o d s ? " Because "how f a r " r e q u i r e s t h a t we consider d e v i a t i o n s b o t h l a r g e r and s m a l l e r t h a n 36, t h e answer t o t h i s q u e s t i o n (again u s i n g Table B) i s 2.093 -1 = t19,.025 $ /fi 7.37 moved r e s i d e n t s " f a r from" 36. 6. What a complicated way t o speak o f p r o b a b i l i t i e s ! The language o f HYPOTHESIS TESTING was developed t o s i m p l i f y p r o b a b i l i s t i c statements such as t h i s . =