You are on page 1of 3

Interpreting SPSS Output Independent Samples t-Test for the Difference between Means Example #1: Do states with

a right-to-work law have lower union membership rate than states without a rightto-work law? 1. HA: States with right-to-work laws have lower union membership rates than states without right-to-work laws. (1 < 2) 2. HO: States with right-to-work laws do not have lower union membership rates than states without rightto-work laws. (1 2) 3. t-test for the difference between two means 4. = 0.01; one-tailed test 5.
Group Statistics STATE HAS RIGHT-TO-WORK 1 0 N 21 30 Mean 13.58 23.10 Std. Deviation 3.700 6.725 Std. Error Mean .807 1.228

% UNION OF EMPLOYED 1982

Independent Samples Test Levene's Test for Equality of Variances t-test for Equality of Means 95% Confidence Interval of the Difference Lower Upper -12.768 -12.472 -6.263 -6.559

F % UNION OF EMPLOYED 1982 Equal variances assumed Equal variances not assumed 10.036

Sig. .003

t -5.880 -6.475

df 49 46.811

Sig. (2-tailed) .000 .000

Mean Difference -9.52 -9.52

Std. Error Difference 1.618 1.470

The 21 states with a right-to-work law (coded as 1) had an average of 13.58% of those employed belonging to a union. In contrast, the 30 states without a right-to-work law had an average of 23.1% of individuals employed as union members. To test for the statistical significance of this difference in union membership between these two groups of states, an independent samples t-test is used. The Levenes F-test for equality of variances equals 10.036 and is statistically significant at the 0.003 level. This indicates that two samples randomly drawn from populations with similar variances would generate an F-test with a value of 10.036 only 3 times out of 1000 trials. This is far less likely than the decision rule of alpha equal to 0.01 (1 time out of 100 trials) that is employed here. Therefore, reject the null hypothesis that the variances of the two populations from which the samples were drawn are equal, and report the t-test not assuming equal variances (second row). The t-test for the difference between these two group means (not assuming equal variances) equals -6.475 and is statistically significant at less than the 0.001 level for a two-tailed test. This is less than a significance value of 0.01 (as indicated by alpha) that is employed here to determine statistical significance. (The precise probability value for this value of t actually is 0.000000052058. This indicates that two samples drawn at random from independent populations with means that are equal 1

will generate a t-test for the difference between the sample means only 5.20 times out of one-hundred million trials. Thus, it is very unlikely that these samples were drawn from populations with equivalent means as assumed by the null hypothesis.) However, in this instance we are employing a one-tailed hypothesis test, so we must divide the reported probability value in the table generated by SPSS by 2. The t-test has a one-tailed probability value of less than 0.001 (its precise one-tailed probability value is 0.000000026029). 6. Test statistics falls inside critical area. 7. Decision: Reject Ho; The difference in union membership rates between states with right-to-work laws and states without right-to-work laws is statistically significant. States with right-to-work laws have statistically significant lower unionization rates than do states without right-to-work laws.

Example #2: Do states with primary enforcement seat belt laws experience a different level of motor vehicle fatalities than states without primary seat belt laws? 1. HA: States with primary enforcement seat belt laws have motor vehicle fatality rates that are different from states without primary enforcement seat belt laws. (1 2) 2. HO: States with primary enforcement seat belt laws have motor vehicle fatality rates that are not different from states without primary enforcement seat belt laws. (1 = 2) 3. t-test for the difference between two means 4. = 0.05; two-tailed test 5.

The 11 states with a primary enforcement seat belt law (coded as 1) had an average of 1.77 motor vehicle fatalities per 100 million vehicle miles traveled (VMT). In comparison, the 40 states without a primary enforcement seat belt law had an average fatality rate of 1.74 motor vehicle fatalities per 100 million VMT. To test for the statistical significance of this difference in state fatality rate between these two groups of states, an independent samples t-test is used. The Levenes F-test for equality of variances equals 1.525 and has a probability value of 0.239, which is not statistically significant at the 95% confidence level. Therefore, we fail to reject the null hypothesis that the variances of the two populations from which the samples were drawn are equal, and report the ttest that assumes equal variances (first row). The independent samples t-test for the difference between these two group means (assuming equal variances) equals 0.235 and has a probability value of 0.815. This probability value is greater than the alpha level of 0.05 that is associated with the 95% confidence level. 6. Test statistics falls outside critical area. 7. Decision: Fail to reject Ho; There is not a statistically significant difference in the motor vehicle fatality rate between states with a primary enforcement seat belt law and states without a primary enforcement seat belt law.

You might also like