In: Statistics and Probability
We would like to test H_0: μ x ≥ μ y in a paired setting. The sample data shows the sample mean of X is larger than the sample mean of Y. And the sample covariance satisfies s_xy >0. Comparing the unpaired two-sample test vs the paired two-sample test, which test shall yield a smaller p-value?
The paired test
Same p-value
Cannot be determined
The unpaired test
Sol:
took
X | Y |
12 | 5 |
20 | 6 |
24 | 7 |
35 | 8 |
use data >data analysis>regression
you will get
t-Test: Paired Two Sample for Means | ||
X | Y | |
Mean | 22.75 | 6.5 |
Variance | 91.58333 | 1.666666667 |
Observations | 4 | 4 |
Pearson Correlation | 0.98478 | |
Hypothesized Mean Difference | 0 | |
df | 3 | |
t Stat | 3.914905 | |
P(T<=t) one-tail | 0.014813 | |
t Critical one-tail | 2.353363 | |
P(T<=t) two-tail | 0.029625 | |
t Critical two-tail | 3.182446 |
p=0.014813(for paired)
t-Test: Two-Sample Assuming Unequal Variances | ||
X | Y | |
Mean | 22.75 | 6.5 |
Variance | 91.58333 | 1.666667 |
Observations | 4 | 4 |
Hypothesized Mean Difference | 0 | |
df | 3 | |
t Stat | 3.365572 | |
P(T<=t) one-tail | 0.021777 | |
t Critical one-tail | 2.353363 | |
P(T<=t) two-tail | 0.043553 | |
t Critical two-tail | 3.182446 |
p=0.021777(for independent t test)
From this example
The paired test yields less p value of 0.0148 compared to 0.021777 for The unpaired test (p value is 0.021777)
ANSWER:
The paired test