Aexd.net

Aexd.net enables the researcher to apply Design of Experiments correctly without having to be an expert in statistics. The user is helped to take critical decisions as to what type of design should be used and how a generated model should be interpreted. Based on each of these decisions, unlogical or incorrect options are hidden such that the user is only presented with information and choices that make common sense in each specific situation.
.
Aexd.net also facilitates generation of statistical reports, tables and figures by automatically generating all data that is of interest for the interpretion of the fitted model in a three steps approach:

1. 1.   creation of experimental design table and experimental results;
   2.   statistical analysis of results with generation of a model and verification if that model would be statistically justified or if the user should proceed with parameters that have the highest impact and go back to step 1;
   3.   dashboard with 3D surfaces and contour diagrams to interactively visualize all interactions of all variables studied.

.
All generated tables and figures can be copied to third-party programs like Microsoft Excel, Word or Powerpoint. All figures, including 3D surfaces, are immediately available in both color and grayscale in high, publication-quality resolution.

Tutorial Aexd.net

3 stages

Easy to start

Explanations & Tip balloons

Integrated into the creation process, this tutorial continues in the form of explanations per section. For example, the first time you start creating an experimental design, you will see an explanation about the fields you need to fill and learn important knowledge that is usefull for that specific section. The next time, the explanation is hidden by default, but you can see it again at any time by clicking the question mark.
.
Pointing your mouse at almost any label, table header, figure title, etc., a small tip balloon will appear. Sometimes, the content of the balloon depends on the actual value of the item that you are pointing on or the result of the analysis. This way, you read valid suggestions that apply to your current, unique situation.

Experimental phase

If the experiments are executed in sequential order (i.e. one by one), it is good practice not to execute the experiments in the same order as in the experimental plan, but to use any random order, especially for the central points. For example, execute one central-point treatment as first experiment, followed by random treatments (including axial points if any), then another central point, followed again by several random treatments, etc.
.
Once all treatments have been executed according to your plan, the results (i.e. the values of the response variables) must be filled into the right-most column.
.
Here is an example of an experimental design plan to discover optimal washing conditions for removal of stains. You can use this example to try out Aexd.net for yourself. To do this, create a design with 3 independent variables, 1 dependent variable, CCRD design, high/low levels: 30-60, 8-30 and 16-60 for ‘Temp’, ‘bleach’ and ‘spin’ respectively. Then, in the Design-and-Results table, fill the response variable (‘stains’) as in the table on the right.

Analysis

Aexd.net offers a very convenient way to perform basic statistical analysis. Once the Design-and-Result table becomes available on the Design page, a new set of buttons appears at the top-right of the page: Design, Analysis and Surface.

For the example below, a screening design had been done before and now, after creating a complete CCRD design with the most significant variables, some relevant statistical analysis results are shown. You can verify that the variables are indeed significant but also which interactions are significant. Moreover, from the presented information you can conclude if it is statistically sound to accept a model that was generated, or if some tweaking is possible or necessary.

Where other DOE softwares allow you to model your process even if it makes no sense statistically (in other words, allow you to produce rubbish science!), Aexd.net warns you how to interprete the analysis and, if possible, explains what to do in case the statistical tests fail for your data.
If you are satisfied with the achieved regression of your model, you will see a mandatory checkbox under the Anova table. Check this, so you can move on to the Surface page.

Surfaces

From the Design or Analysis page, go to the Surface page by clicking the ‘Surface’ button on the upper-right corner of the screen. Aexd.net now automatically generates interactive 3D surface plots and contour diagrams for every possible combination of two independent variables. Click on a surface plot, hold and move your mouse to see the 3D surface from different angles. Below are some static images (non-interactive) shown as examples.

Each plot shows the dependance of the response variable (here ‘stains’ on the vertical axis) to two independent varables (each on one of the horizontal axes that define the bottom of the surface, e.g. ‘spin’ and ‘bleach’). In the contour plot, these latter are the x and y axes of the plot. Thus, if you have three independent variables in your model in stead of two, one independent variable cannot appear in the plot and is fixed at a certain value. A ‘fixed variable’ is then called ‘parameter’. In such case, or if there are even more parameters, a tool appears as in the image below, showing the value of the variables and parmeter(s) in white. When you move your mouse over the tool you will see sliders appear. You can move the sliders to see interactively how changing the value of a parameter affects the 3D surface and the contour plot. This makes it very convenient to learn and understand the mechanism of your process and find an optimum, minimum or desired value.

Where are my designs?

Frequently Asked Questions (FAQ)