In: Statistics and Probability
Ten randomly selected people took an IQ test A, and next day they took a very similar IQ test B. Their scores are shown in the table below.
Person | A | B | C | D | E | F | G | H | I | J |
Test A | 99 | 86 | 112 | 90 | 106 | 78 | 120 | 95 | 118 | 101 |
Test B | 102 | 86 | 109 | 92 | 104 | 79 | 121 | 97 | 115 | 106 |
1. Consider (Test A - Test B). Use a 0.01 significance level to
test the claim that people do better on the second test than they
do on the first. (Note: You may wish to use software.)
(a) What test method should be used?
A. Matched Pairs
B. Two Sample z
C. Two Sample t
(b) The test statistic is
(c) The critical value is
(d) Is there sufficient evidence to support the claim that
people do better on the second test?
A. No
B. Yes
2. Construct a 99% confidence interval for the mean of the
differences. Again, use (Test A - Test B).
_____<μ<____
Let be mean test score for Test A
and be mean test score for Test B
Here we have to test
a) Here we have to use two sample t test since two samples are independent.
But first we have to check population variances are equal or not.For that we have to use F test
Test statistic:
P value from excel using command:
=FDIST(1.138,9,9) (Degrees of freedom for numerator = n1-1=10-1= 9 and degrees of freedom
=0.4252 for denominator =n2-1=10-1 = 9)
P value = 2*0.4252 = 0.8505
Here P value > alpha = 0.01
So we do not reject H0
That means population variances are equal
So we can use here pooled two sample t test
b) Test statistic:
where
= 13.3535 (Round to 4 decimal)
= 5.9719 (Round to 4 decimal)
t = -0.100 (Round to 3 decimal)
Test statistic = -0.100
c) Critical value:
degrees of freedom = n1+n2 - 2 = 10 + 10 -2 = 18
alpha = level of significance = 0.01
Critical value = 2.552 (From statistical table of t values)
Here |t| = 0.1 < critical value = 2.552
So we do not reject H0
d) No.There is not sufficient evidence to support the claim that people do better on the second test
2) 99% confidence interval for is
Where tc is t critical for degrees of freedom = n1+n2-2 = 10+10-2 = 18 and confidence level = c= 0.99
tc = 2.878 (From statistical table of t values)
99% confidence interval for mean of differences is (-17.7871 , 16.5871)