Technical Support – Help
Frequently Asked Questions
These are the most commonly asked questions about DxTER. Please feel free to contact us in case you don’t find here the answer you are looking for.
An organization in DxTER represents your company, university, or research unit. It serves to group users, manage permissions, and ensure data isolation between different entities using the platform.
You have several options for uploading your experimental data to DxTER:
- Import data from Excel files
- Import data from CSV files
When importing from Excel or CSV, it’s important to use flat files (single table format with rows and columns).
Here’s an example of how the table should look:
| Temperature (°C) | Pressure (atm) | Catalyst Amount (g) | Expected Yield (%) | Expected Purity (%) |
| 80 | 2.5 | 0.5 | 75 | 97 |
| 95 | 3.0 | 0.8 | 80 | 87 |
Your administrator will either invite you via email. Follow the instructions in the email to complete your setup. If you don’t receive your invitation, please, talk with your organization administrator.
To create a dashboard:
- Navigate to the “Data Analysis” section in DxTER.
- Click on “New Dashboard”
- Give your dashboard a name and select a layout template (number of columns).
- Add widgets to your dashboard by clicking “Add Widget” and selecting the type of visualization you want:
- Charts: Lines, Bars, Scatter
- Double-y charts: Lines, Scatter
- Pareto front
- Correlation matrix
- Configure each widget by selecting the relevant Factors and Responses and customizing the display options.
- Arrange the widgets on the dashboard by changing the layout (number of columns).
- The dashboard will autosave as you make changes.
The table generated by DxTER presents a set of experiment recommendations designed to help you optimize your experiment project and reduce the number of steps needed to achieve the desired result.
Each row in the recommendation table represents a proposed experiment, with specific settings for each factor as determined by DxTER’s analysis. These factor settings are carefully chosen to explore the experimental space efficiently and reveal important relationships between factors and responses.
For each recommended experiment, DxTER provides:
- Factor settings (or “Factor to be tested”): The specific values or levels at which each factor should be set for the experiment.
- Expected response ranges (or “Expected values”): The anticipated ranges for each response variable, based on the model’s predictions.
To illustrate this, let’s consider a hypothetical DOE example for optimizing a chemical reaction:
| Factors (To be tested) | Responses (Expected Values) | ||||
| Temperature (°C) | Pressure (atm) | Catalyst Amount (g) | Expected Yield (%) | Expected Purity (%) | |
| Rec. 1 | 80 | 2.5 | 0.5 | 75 (+/- 5%) | 97 (+/- 1.5%) |
| Rec. 2 | 95 | 3.0 | 0.8 | 80 (+/- 7%) | 87 (+/- 6%) |
In this example, DxTER recommends two experiments. For the first experiment, you would set the temperature to 80°C, pressure to 2.5 atm, and use 0.5 g of catalyst. The expected yield should fall between the confidence interval 75% +/- 5%: (i.e., between (70% – 80%)), with a purity between the confidence interval 97% (+/- 1.5%) (i.e., between (95.5% – 98.5%)).
By conducting these recommended experiments and comparing the actual results to the expected ranges, you can:
- Validate or refine the underlying model.
- Identify optimal factor settings.
- Discover unexpected interactions between factors.
Remember, these recommendations are based on statistical analysis and should be used in conjunction with your domain expertise.
Filters: These are variables that are not used in the experimental design. They are typically nominal dimensions that serve to label experiments, such as the experiment author, execution date, or metadata-related information. DxTER allows these variables to be used for filtering the project dataset, selecting experiments in a DOE, or plotting series in graphical widgets.
Factors: These are the variables in an experiment that influence the outcome. Factors can be numerical (e.g., temperature, pressure, concentration) or categorical (e.g., types of solvents, material sources). They serve as inputs to the experimental process and are chosen based on their relevance to the desired outcomes.
Responses: These are the measurable outputs of the experiment. Responses provide insights into how factors affect the system being studied. Examples include the yield of a chemical reaction, the optical properties of a material, or the durability of a product under specific conditions.
If you’re an administrator, go to the user management section of the application and invite users by entering their email addresses.
Dashboards in DxTER are user specific. They are not shared at the project level, meaning each user has their own individual dashboards.
This design ensures that every researcher can create and customize their dashboards independently, analyzing experiments in a way that suits their specific requirements. Since creating and configuring widgets is quick and straightforward, users can adapt their dashboards as needed without impacting others.
Each time you request recommendations from DxTER, you can specify the number of experimental suggestions you need. Keep in mind that requesting more recommendations increases computational cost and may extend the time required for DxTER to generate results.
Additionally, the number of objectives (responses to optimize) exponentially increases computational complexity. Therefore, DxTER sets certain limits to ensure the process remains efficient and manageable.
- Single-objective optimization (1 response): You can select a batch of up to 20 experimental recommendations.
- Multi-objective optimization (2 or 3 responses): You can select a batch of up to 5 experimental recommendations.
In a project, if you want to add new experiments, you need to do it manually by adding a row for each experiment to the data table.
Adding new experimental results is a key part of using DxTER effectively. After receiving recommendations and conducting new experiments, you can incorporate those results to obtain more precise recommendations.
Please note that, at this moment, once the project is created, we do not allow users to add multiple experiments via an Excel/CSV file to maintain data integrity.
Every organization has, at least, an administrator in DxTER who is responsible for managing users, assigning roles, and approving new registrations within your organization.
DxTER uses Bayesian optimization, a probabilistic technique, to identify the best experiments to conduct based on your DOE configuration. This approach is inherently non-deterministic, meaning that even with the same data and setup, the recommendations may vary slightly between iterations.
This variability arises because Bayesian optimization explores the experimental space probabilistically to find optimal solutions, and explains why you would get different results when applying the same DOE to the same data, even without updates or new experiments.
You can modify any data for existing experiments at any time. Remember that you can also use the field “Note” if you want to add any contextual information, description or metadata about a given experiment.
DxTER allows you to create and configure an unlimited number of dashboards. There are no restrictions on the number of dashboards or the number of widgets you can add to each dashboard.
This flexibility enables you to organize your dashboards based on your specific needs, such as by type of analysis, experimental factors, or any other criteria that help you study, analyze, and interpret your experimental data effectively.
DxTER’s policy on multiple organization membership is as follows:
- You can only use the same email address in one organization
- If you need to join multiple organizations, you must use different email accounts for each one.
This policy ensures clear separation between organizations and maintains data integrity within the DxTER platform.
You can ask DxTER for recommendations as many times as needed to achieve your desired results. Each time you request new recommendations, DxTER applies the DOE configuration you have set up to your experimental dataset.
If you have added new experiments or modified existing ones, DxTER will automatically take these updates into account.
It’s normal to have missing data for some variables in experiments. For generating recommendations, if a variable with missing data is part of the DOE definition, DxTER will automatically ignore that experiments.
Experiments with incomplete data can still be used in (Visual) Data Analysis.
Sometimes, it’s not possible to use line or bar charts in DxTER due to visualization restrictions. This typically happens when you’ve selected two variables to plot, and there are multiple Y-axis values for a single X-axis value. In this cases, it becomes impossible to accurately represent this data using lines or bars.
Example: Consider plotting Temperature (X-axis) against Resistance (Y-axis) with the following data:
- Experiment 1: (30°C, 10mV)
- Experiment 2: (30°C, 15mV)
In this case, for the X value of 30°C, we have two different Y values (10mV and 15mV). A line or bar chart can’t properly represent this scenario.
To handle these situations, DxTER automatically switches to a scatter plot and provides you with a caution message. This ensures all your data points are represented without any loss of information or accuracy.
The Pareto Front widget in DxTER is a powerful visualization tool that helps you make sense of your experimental data when you’re trying to optimize multiple objectives at once. Here’s what it does for you:
- Shows you the best possible outcomes (Pareto-optimal solutions) when you’re balancing two different goals in your experiments.
- Lets you select two objectives you want to optimize – these are typically Responses you want to maximize or minimize.
- Displays the trade-offs between your chosen objectives in an easy-to-understand graph.
- Helps you spot which experiments give you the best results across both objectives.
This tool is particularly handy when you’re dealing with competing goals and need to find the sweet spot. For example, if you’re trying to maximize product quality while minimizing production cost, the Pareto Front widget will show you the range of optimal solutions.
By using this widget, you can:
- Quickly identify your most promising experimental results
- Understand how improving one objective affects another
- Make informed decisions about your next steps in the experimental process
The Correlation Matrix in DxTER provides a comprehensive view of relationships between all Factors and Responses in your experimental dataset. Here’s what you need to know:
- The matrix includes all Responses and Factors from your experiment dataset
- There is no option to exclude specific Factors or Responses from the calculation
However, while you can’t filter out specific variables, DxTER does offer flexibility in how you view the data:
- You can adjust the range of linear correlation coefficients you want to see
- This range can be set anywhere between -1 and 1
For example, you might choose to display only strong correlations by setting a range of 0.7 to 1 or -1 to -0.7. This helps you focus on the most significant relationships in your data.
