Small issues

When installing the TVMM package, type the following command: install.packages(“TVMM”, dependencies = TRUE). If you have already installed the package when trying to use the interactive guide (guiTVMM ()), a message will appear. Install the tkrplot package as the following command line: install.packages(“tkrplot”).

The gui of TVMM package

This GUI aims to offer the user an interactive way to carry out decision-making procedures under the vector of population means of non-normal, normal, and normal contaminated populations. For this, the TVMM package available in software R was built with the main objective of providing alternative statistical tests to the well-known T2 test of Hotelling that works for high-dimensional data and are more general, not requiring multivariate normality of the data. One of the limitations of the traditional T2 test is that it is not applicable for data of this nature due to the singularity of the covariance matrix. In addition, it should not be used under non-normality and heterogeneity of the variances. Our proposed tests adapted to the likelihood ratio test (LRT) do not present these limitations.

How to prepare the data?

The data must be saved in an excel file with the extension .csv or a notepad file with extension .txt, one file for the data matrix and another for the vector of averages (see Figure 1).

Figure 1: Data presentation: observation matrix and vector of means (.txt).

How to use the GUI?

To use this GUI it is necessary to have software R (https://cran.r-project.org/) and RStudio, free version (https://rstudio.com/products/rstudio/download/). Then install and load the TVMM package. Right after, type the following command: guiTVMM(gui = TRUE). A graphical interface will appear, as shown in Figure 2.

Figure 2: Installing the package and loading the GUI.

The menu File options

In this menu we have the options: Data, View / Edit and Quit. The Figure 3 details the File menu. In the Data option we can load the data set. A window will open and the user will be able to search, on his machine, where this data is saved and load it in R . This can be done in both the csv and txt options, as shown in Figure 1. The user is allowed to view and edit the data. To do so, just click on the View / Edit option. Then, the user will make his choice of what he wants to see or edit: the observation matrix or the vector of averages. We remind you that this option will only be available after the user loads the data into R. Finally, this GUI can be finalized by clicking on the Quit button, select yes or no. If you choose yes, the GUI will automatically close. Otherwise, nothing will happen.

Figure 3: The menu File: Data; View/edit and Quit.

The menu Tests options

The next menu is the Tests menu, where the user has the select option. By clicking on select, new options emerged, which are the six tests presented in the TVMM package and which can be consulted in Alves and Ferreira (2019). Just click on these options and a window will open containing a message with the value of the test statistic and the associated p-value. We consider the level of significance to be 5%, which is usual. But this does not prevent the user from considering the other two levels of significance: 1% and 10%. They are not used much. But the user has that freedom (see Figure 4). We also remember that only the likelihood ratio (LRT) tests adapted with dashes (LRTT) are valid for high-dimensional data.

Figure 4: The menu tests: the T2 traditional test.

For parametric bootstrap versions, the user has the option to choose the number of bootstrap simulations. By default, this is equal to 2,000. We recommend using at least 2,000 simulations as, for example, in the T2Boot test (see Figure 5). As a default, we set B = 80, so that the examples don’t take too long to run.

Figure 5: The menu tests: the T2Boot test.

The menu Graphs options

We also inserted a graphical output option (histograms) that illustrate the decision making on the hypothetical means vector (acceptance or rejection of the null hypothesis) for each of these tests. This can be seen in the Graphs menu. In this menu, you have the options for the graphical display of all tests available in the TVMM package. We remember that only the adapted versions of the Likelihood Ratio (LRT) test are valid for high-dimensional data (p > n). We will exemplify only for the traditional T2 test. For the other tests, the procedure is analogous. When clicking on the original T2 option, a graphic window will open containing the question: do you want to view this fugures in color?. If the answer is yes, a window will open containing two colored histograms that represent the associated probability density function, the critical region and the observed statistic, respectively (see Figure 6). If the answer is no, the same histograms are presented in gray scale (see Figure 7). Besides, the 95% quantile values and the observed statistic are presented. We also present the option to save this output on your machine. Note that there is a Save option in four formats: png, jpeg, pdf, eps, bmp and tiff.

Figure 6: The menu Graphs: the T2 test graphs in colors.

Figure 7: The menu Graphs: the T2 test graphs in colors.

The menu Help option

Finally, we present the Help menu that when selecting this Manual option (Figure 8). This Help menu also presents help options with the testing methodology and the hypothesis presented here.

Figure 8: The menu Help: the options of the Help menu..

Alves, Henrique J. P., and Daniel F. Ferreira. 2019. “Proposition of New Alternative Tests Adapted to the Traditional \(T^2\) Test.” Communications in Statistics – Simulation and Computation. https://doi.org/https://doi.org/10.1080/03610918.2019.1693596.