Using Parallel Feature Analysis (PFA), NXG Logic Explorer can simultaneously perform t-tests, Mann-Whitney tests, and Bartlett's homogeneity of variance tests for multiple features, including multiway contingency table analysis with chi-squared tests of association for multiple categorical variables across the same two groups. Thus, many parametric and non-parametric 2-sample tests for continuous and categorical features are performed, which typically takes less than a minute. Any time Bartlett's test is significant for a continuously-scaled feature, the degrees of freedom of the t-statistic is automatically changed.
The output involves a single table for which fonts and colors can be changed, based on the needs of the user -- using a pastable html or graphic image.
All of the above tests can be employed simultaneously for 20-30 features in less than minute.
Step 1: Start Explorer, and input the Retinol dataset (5 seconds).
Step 2: Select independence tests, unpaired 2-sample (2 seconds).
Step 3: Select SEX as the 2-group variable, and the remaining continuous and categorical variables (5 seconds).
Step 4: Select the Grouped analysis icon and view the output table (5 seconds). Notice the adjustment to the d.f. of the t-test when Bartlett's test is significant.
This implies that the Welch t-test for unequal variances was used. The categorical tests for vitamin use and smoking provide the asymptotic p-values along with
the Fisher exact test based on a set limit of 10,000 iterations. This can be increased to 100,000 if necessary, but 10,000 has shown to converge on results from 100,000 iterations.
Chi-squared tests are run the same time parametric and non-parametric tests are run on the continuous variables.
In summary, the total time required to run the 4 steps was less than a minute, and the results included a dozen variables and dozens of tests, while all the necessary tests concerning equality of variances were ran behind the scenes, with Welch tests employed if heteroscedasticity was present. The chi-squared tests were also performed on the categorical variables which happened to be included in the 2-group tests.
The above runs are ideal for generating demographic tables for intervention studies, case-control, cohort and prospective, and baseline variables for randomized clinical trials and longitudinal studies with k-treatment groups. Group labels can be easily input using the label editor or by input of text files containing categorical group labels.
The analyses presented above are also available in the free Standard edition of Explorer. Once you become accustomed to the time-savings offered by using Explorer, you will begin to realize how much time you have wasted with your current software. We also use Explorer whenever possible, since we can't afford to waste time and lose the productivity associated with use of so many other platforms.
