In: Statistics and Probability
The situation…
• A 12-month email study looked at the click-through responses triggered by two versions of email, Version A and Version B.
• In all 24 distributions (12 for Version A; 12, Version B) emails were drawn randomly.
• People receiving A or B versions were unrelated (randomly selected every time for both version of the email).
• The researchers had no idea which version would trigger more clicks.
• Alpha was set at p <.05.
MONTH | Group A | Group B |
January | 36 | 36 |
February | 35 | 25 |
March | 24 | 29 |
April | 23 | 37 |
May | 29 | 37 |
June | 12 | 23 |
July | 18 | 27 |
August | 12 | 21 |
September | 14 | 24 |
October | 24 | 33 |
November | 25 | 32 |
December | 33 | 40 |
Mean | ||
SD | ||
Variance | ||
t test | ||
Effect Size | ||
MONTH | Group A | Group B |
January | 36 | 36 |
February | 35 | 25 |
March | 24 | 29 |
April | 23 | 37 |
May | 29 | 37 |
June | 12 | 23 |
July | 18 | 27 |
August | 12 | 21 |
September | 14 | 24 |
October | 24 | 33 |
November | 25 | 32 |
December | 33 | 40 |
Mean | 23.75 | 30.33333 |
Standard Deviation | 8.508018 | 6.372288 |
Two sample t test is used to compare the means assuming equal variance. The test is performed in following steps,
Step 1: The Null and Alternative Hypotheses
Step 2: Select the appropriate test statistic and level of significance.
The t statistic is used to compare the two population means and the significance level is for the test (Generally 5% significance level used to compare two means)
Step 3: State the decision rules.
The decision rules state the conditions that if,
,
Step 4: Compute the appropriate test statistic and make the decision.
The t statistic is used to compare the two population means. The t statistic is computed using the formula,
The P-value for the t statistic is obtained using the t distribution table for degree of freedom = n -1 = 24 -1 = 23,
Decision:
The corresponding P-value is 0.06 which is less than 0.05 at the 5% significance level for the two sided alternative hypothesis.
Step 5: State the Conclusion
The null hypothesis is rejected. Hence there is a statistically significant evidence that the Group 1 has less click-through responses ( t value is negative)
The effect size is obtained by using the formula,
Cohen's d value, d=0.2 considered a 'small' effect size, 0.5 represents a 'medium' effect sizeand 0.8 a 'large' effect size