๐ : THE ONE-SAMPLE ๐ข -TEST AND SPSS Business Statistics
CONTENTS The one-sample ๐ข -test for ๐ Hypotheses and SPSS Old exam question Further study
THE ONE-SAMPLE ๐ข -TEST FOR ๐ When to use the one-sample ๐จ -test? โช It relies on the fact that the distribution of เดค ๐ is normal with ๐ ๐ ๐ = ๐ ๐ and ๐ เดค ๐ , so on the CLT ๐ เดค ๐ = โช So it can be used to test a hypothesis of the mean only โช not a hypothesis on ๐ , median, etc. โช It works when the population ๐ is normal โช or when ๐ is symmetric and ๐ โฅ 15 โช or when ๐ โฅ 30 โช It requires knowledge of ๐ ๐ Rule of thumb: โช so knowing ๐ก ๐ will not work use of CLT for mean justified when ๐ โฅ 30 โช that is a problem!
THE ONE-SAMPLE ๐ข -TEST FOR ๐ So, what to do if you donโt know ๐ ๐ ? 2 from the โช Luckily, we can estimate the population variance ๐ ๐ 2 sample variance ๐ก ๐ เดค เดค เดค เดค ๐โ๐ เดฅ ๐โ๐ เดฅ ๐โ๐ ๐ ๐โ๐ ๐ โช calculate the test statistic ๐ก ๐ / ๐ instead of ๐ ๐ / ๐ ? ๐ ๐ = = ๐ก เดฅ ๐ เดฅ ๐ ๐ Sure, we can ... เดค ๐โ๐ เดฅ โช but instead of ๐ ~๐ 0,1 , we have ๐ ๐ เดฅ เดค ๐ โ ๐ เดค ๐ ~๐ข ๐โ1 ๐ก เดค ๐ โช where ๐ข df is the ๐ข -distribution with df degrees of freedom
THE ONE-SAMPLE ๐ข -TEST FOR ๐ Example: โช sample of ๐ = 100 body heights โช sample mean าง ๐ฆ = 179.1 cm 2 = 212.4 cm 2 โช sample variance ๐ก ๐ We think (or hope, or fear) that ๐ ๐ < 181 cm Use the five-step procedure โช Steps 1-2: as before ( ๐ผ 0 : ๐ โฅ 181 ) โช Steps 3-5: see below
THE ONE-SAMPLE ๐ข -TEST FOR ๐ โช Step 3 เดค ๐โ181 โช under least extreme version of ๐ผ 0 : 212.4/100 ~๐ข 99 โช no further assumptions required (because ๐ โฅ 30 ) โช Step 4 โช value of test statistic ๐ข calc = โ1.3037 โช critical value of ๐ข from the table is ๐ข crit,lower,0.05,df=99 = โ 1.660 โ1.660 0 โช Step 5 โช ๐ข calc โ ๐ crit , so do not reject ๐ผ 0 โช conclude that there is no evidence for ๐ < 181
THE ONE-SAMPLE ๐ข -TEST FOR ๐ Summing up: โช ๐ ๐ known: โช use ๐จ -table/perform ๐จ -test โช ๐ ๐ estimated by ๐ก ๐ : โช use ๐ข -table/perform ๐ข -test with df = ๐ โ 1 degrees of freedom โช In both cases: โช use two-tailed critical value ( ๐ฝ/2 ) to construct a confidence interval โช use two-tailed critical value ( ๐ฝ/2 ) to construct a critical region for a two-sided hypothesis test โช use one-tailed critical value ( ๐ฝ ) to construct a critical region for a one-sided hypothesis test
EXERCISE 1 We test a hypothesis ๐ผ 0 : ๐ = 310 with a sample size ๐ = 50 and ๐ unknown. Which statements are correct? A. We can use the ๐จ -test. B. We can use the ๐ข -test. C. We canโt perform this test. D. The sample is normally distributed. E. The population is normally distributed. F. The sampling distribution of the mean is normal.
HYPOTHESES AND SPSS Using SPSS for the one-sample ๐จ -test โช This is not possible! โช SPSS (realistically) assumes you donโt know ๐ ๐ โช SPSS will always use ๐ก ๐ to estimate ๐ ๐ and then do a one-sample ๐ข -test
HYPOTHESES AND SPSS ๐ =life Using SPSS for the one-sample ๐ข -test expectancy โช Claim: the average life expectancy is 68 year ๐ผ 0 : ๐ ๐ = 68 Under ๐ผ 0 : ๐~๐ข 152 ๐ -value ๐ข calc
HYPOTHESES AND SPSS But look carefully: โช it is a two-sided test (SPSS says: two-tailed) ๐ผ 0 : ๐ ๐ = 68 ๐ -value
HYPOTHESES AND SPSS Doing a one-sided test with SPSS โช How to do a one-sided ๐ข -test in SPSS? โช we could do it by hand with the intermediate results โช we could do it by hand with the two-sided results โช you must be able to do both โช Example: โช null hypothesis ๐ผ 0 : ๐ โฅ 68 โช sample with ๐ = 153 yields าง ๐ฆ = 64.515 and ๐ก = 12.7937 64.515 โ 68 153 โ 1
HYPOTHESES AND SPSS 68 + 68 โ 64.515 Two-sided โช ๐ เดค ๐ โค 64.515 + ๐ เดค ๐ โฅ 71.485 = ๐ แ ๐ข โค 64.515โ68 71.485โ68 12.7937/ 153 = ๐แบ 12.7937/ 153 + ๐ ๐ข โฅ แ ๐ข โค แป โ3.369 + ๐ ๐ข โฅ 3.369 โช use SPSS for ๐ข df=152 : ๐ = 0.001
HYPOTHESES AND SPSS One-sided (right-sided) and าง ๐ฆ < ๐ 0 64.515โ68 โช ๐ เดค ๐ โค 64.515 = ๐ ๐ข โค 12.7937/ 153 = ๐ ๐ข โค โ3.369 โช this is exactly half of the reported two-sided ๐ -value โช ๐โvalue = 0.0005 So, to move from the SPSS-reported two-sided ๐ -value to right-sided ๐ -value, in case of าง ๐ฆ < ๐ 0 : โช divide by 2 ๐ right = ๐ two /2
HYPOTHESES AND SPSS One-sided (left-sided) and าง ๐ฆ < ๐ 64.515โ68 โช ๐ เดค ๐ โฅ 64.515 = ๐ ๐ข โฅ 12.7937/ 153 = ๐ ๐ข โฅ โ3.369 โช this is exactly 1 โ ๐ ๐ข โค โ3.369 โช ๐โvalue = 0.9995 So, to move from the SPSS-reported two-sided ๐ -value to left- sided ๐ -value, in case of าง ๐ฆ > ๐ 0 : โช divide by 2 and subtract the result from 1 ๐ left = 1 โ ๐ two /2
าง HYPOTHESES AND SPSS Another example โช Suppose โช ๐ผ 0 : ๐ = 3 โช ๐ฝ = 0.05 โช ๐ฆ = 2.84 ๐=3 เดค โช ๐โvalue = 2 ร ๐ ๐ โค 2.84 = 0.0456
HYPOTHESES AND SPSS Both าง ๐ฆ = 2.84 and าง ๐ฆ = 3.16 give the same two-sided ๐ - value! โช so which is true when ๐ผ 0 : ๐ โฅ 3 and which when ๐ผ 0 : ๐ โค 3 ?
าง าง HYPOTHESES AND SPSS ๐ฐ ๐ : ๐ โค ๐ ๐ฐ ๐ : ๐ โฅ ๐ (right-sided) (left-sided) ๐ SPSS ๐ SPSS ๐ฆ = 2.84 ๐=3 เดค ๐=3 เดค ๐ ๐ โฅ 2.84 = 1 โ ๐ ๐ โค 2.84 = 2 2 ( าง ๐ฆ < ๐ 0 ) ๐ SPSS ๐ SPSS ๐=3 เดค ๐=3 เดค ๐ฆ = 3.16 ๐ ๐ โฅ 3.16 = ๐ ๐ โค 3.16 = 1 โ 2 2 ( าง ๐ฆ > ๐ 0 )
EXERCISE 2 Suppose we test a hypothesis on the mean ๐ผ 0 : ๐ โฅ 310 with significance level ๐ฝ = 0.05 . We sample data, and perform the test and find a sample mean าง ๐ฆ = 307 and a two- sided ๐ -value 0.08 . What do we conclude?
EXERCISE 3 Suppose we have the following result: A. reject ๐ผ 0 : ๐ โค 6 . B. ๐โvalue < ๐ฝ . C. probably ๐ > 6 . What is a correct sequence? AโBโC , CโBโA , BโAโC , etc?
OLD EXAM QUESTION 26 March 2015, Q3b
FURTHER STUDY Doane & Seward 5/E 9.1-9.5 Tutorial exercises week 2 one-sided test in SPSS
Recommend
More recommend