1. 1 Hypothesis Testing Basic Problem We are interested in deciding whether some data credits or discredits some “hypothesis” (often a statement about the value of a parameter or the relationship among parameters).

2. 2 Suppose we consider the value of  = (true) average lifetime of some battery of a certain cell size and for a specified usage. and hypothesize: H 0 :  = 160 H 1 :  160

3. 3 This would usually involve a scenario of either (1) 160 is a “standard” (2) the previous  was 160 (has there been a change?) or something analogous. H 0 :  = 160 H 1 :  160 H 0 is called the “Null Hypothesis” H 1 is called the “Alternate Hypothesis”

4. 4 We must make a decision whether to ACCEPT H 0 or REJECT H 0 (ACCEPT H 0 same as REJECT H 1 ) (REJECT H 0 same as ACCEPT H 1 ) We decide by looking at X from a sample of size n

5. 5 Basic Logic: (1)Assume (for the moment) that H 0 true. (2)Find the probability that X would be “that far away”, if, indeed, H 0 is true. (3)If the probability is small, we reject H 0 ; if it isn’t too small, we give the benefit of the doubt to H 0 and accept H 0. H 0 :  = 160 H 1 :  160

6. 6 BUT — what’s “too small” or “not too small”? Essentially — you decide! You pick (somewhat arbitrarily) a value, , usually.01 .10, and most often  =.05, called the SIGNIFICANCE LEVEL;

7. 7 If the probability of getting “as far away” from the H 0 alleged value as we indeed got is greater than or equal to , we say “the chance of getting what we got isn’t that small and the difference could well be due to sample error, and, hence, we accept H 0 (or, at least, do not reject H 0 ).”

8. 8 If the probability is < , we say that the chance of getting the result we got is too small (beyond a reasonable doubt) to have been simply “sample error,” and hence, we REJECT H 0.

9. 9 Suppose we want to decide if a coin is fair. We flip it 100 times. H 0 : p = 1/2, coin is fair H 1 : p 1/2, coin is not fair

10. 10 Let X = number of heads Case 1)X = 49 Perfectly consistent with H 0, Could easily happen if p = 1/2; ACCEPT H 0 2)X = 81 Are you kiddin’? If p = 1/2, the chance of gettin’ what we got is one in a billion! REJECT H 0 3)X = 60 NOT CLEAR!

11. 11 What is the chance that if p = 1/2 we’d get “as much as” 10 away from the ideal (of 50 out of 100)? If this chance < , reject H 0 If this chance > , accept H 0

12. 12 Important logic: H 0 gets a huge “Favor from the Error”; H 1 has the “Burden of Proof”; We reject H 0 only if the results are “overwhelming”.

13. 13 To tie together the  value chosen and the X values which lead to accepting (or rejecting) H 0, we must figure out the probability law of X if H 0 is true. Assuming a NORMAL distribution (and the Central Limit Theorem suggests that this is overwhelmingly likely to be true), the answer is: X  = 160

14. 14 We can find (using normal distribution tables) a region such that  = the probability of being outside the region:  /2 150.2  =160 169.8  /2 X (I made up the values of 150.2 and 169.8)

15. 15 Note:logic suggests (in this example) a “rejection” region which is 2- sided; in experimental design, most regions are 1-sided. 150.2 169.8is called the Acceptance Region (AR) 169.8is called the Critical Region (CR)

16. 16 Decision Rule: If X in AR, accept H 0 If X in CR, reject H 0  /2 150.2  =160 169.8  /2 X

17. 17 X is called the “TEST STATISTIC” (that function of the data whose value we examine to see if it’s in AR or CR.)  20C   Critical Value H 0 :  > 20 H 1 :  < 20 ONE-SIDED LOWER TAIL H 0 :  < 10 H 1 :  >10  X 10 C  Critical Value ONE-SIDED UPPER TAIL X

18. 18  has another meaning, which in many contexts is important: Good! Good!(Correct!) H 0 trueH 0 false Type II Error, or “  Error” Type I Error, or “  Error” Good! Good!(Correct) we accept H 0 we reject H 0

19. 19  =Probability of Type I error = P(rej. H 0 |H 0 true)  =Probability of Type II error = P(acc. H 0 |H 0 false)

20. 20 We often preset . The value of  depends on the specifics of the H 1 : (and most often in the real world, we don’t know these specifics).

21. 21  C=14 1 EXAMPLE: H 0 :  < 100 H 1 :  >100 Suppose the Critical Value = 141:  =100 X

22.  = P (X < 141|H 0 false) 141 What is  ?  = P (X < 141/  = 150).3594 =.3594  = 150 141  = P (X < 141/  = 160).2236 =.2236  = 160 141  = P (X < 141/  = 170).1230 =.1230  = 170 141  = P (X < 141/  = 180)  = 180 These are values corresp.to a value of 25 for the Std. Dev. of X =.0594

23. 23 Note:Had  been preset at.025 (instead of.05), C would have been 149 (and  would be larger); had  been preset at.10, C would have been 132 and  would be smaller.  and  “trade off”.

24. 24 In ANOVA, we have H 0 :  1  2  =  c H 1 : not all (column) means are =. The probability law of “F calc ” in the ANOVA table is an F distribution with appropriate degrees of freedom values, assuming H 0 is true: 0  C  Critical Value

25. 25 F calc = MSB col MSW Error E(MSB col ) =  2 + V col E(MSW Error ) =  2 The larger the ratio, F calc, the more suggestive that H 0 is false).  C C is the value so that if V col = 0 (all  ’s=) P (F calc > C)= 

26. 26 Note: What is  ?  = P < C F calc ( The  ’s are not all = (i.e., the level of the factor does matter!!) ) Answer: Unable to be determined because we would need exact specification of the “non-equality”. [Hardly ever known, in practice!]

27. 27 HOWEVER — The fact that we cannot compute the numerical value of  in no way means it doesn’t exist! And – we can prove that whatever  is, it still “trades off” with .