Parametric & Nonparametric Testing

In our third attempt to support the hypothesis that: “Married females make more than non-married females” a non-parametric means of testing the data was used. In the two previous hypothesis-testing papers parametric tests such as a one and two-tailed Z test were used. The null hypothesis is that earnings for southern women are dependent on marital status. The alternate hypothesis is earnings for southern women are not dependent on marital status. The numerical hypothesis statement for the null and alternate hypothesis is whether expected chi square of $13,275.72 is different than the observed chi square.

Although effective, the Z test does not give as much detail to the hypothesis or expected result as the Chi-square test does. The chi-square test has an expected value that is used to compare to the actual observed value during the test. In the hypothesis “Married females make more than non-married females”, the expected value using chi for southern married females is 222592.55 in one sample and 299088.45 in the second sample. These numbers reflect a fairly even population and therefore the result of the test should prove accurate.

Once each of the samples is calculated, the non-parametric test shows once again that the hypothesis is incorrect. The data supports the fact the non-married females make more than married females and not by a small margin. The final decision is to once again reject the Ho: “Married females make more than non-married females”.

This five step process allows the research a checks and balance opportunity. During the first paper the five step process proved tedious and overwhelming. After completing the five step test for the third time, finally an appreciation for the flow and necessity of the test has been reached by the research team. Below is the data we used to analyze during the nonparametric test.

s/nm s/m ns/nm ns/m

15234.00 19306.00 21716.00 29977.00

16817.00 13318.00 22133.00 33959.00

20793.00 32094.00...