- How can the EDA help you?
- Improve your experiment with tailored feedback
- Find statistical analysis methods appropriate for your experimental design
- Get help determining your sample size
- Generate a randomisation sequence taking into account features of your experiment
- Share key features of your experiment using an automatically generated PDF
- Share your experiment using a URL of a read-only version of an EDA diagram
- Keep up to date with the EDA
The EDA is a free-to-use online tool to help you design in vivo experiments more rigorously. It provides guidance as you design your experiments, and helps you communicate plans clearly with colleagues, collaborators, funders, ethical review bodies and in publications.
Experiments designed in the EDA are machine-readable. Computer-based logical reasoning determines what feedback is appropriate for each experimental plan. What feedback is given, and when, was determined by a working group of experts in experimental design and statistics.
The EDA has a function called Critique that gives feedback tailored to your experiment. The feedback covers common ways to improve experimental design, such as prompting you to consider implementing randomisation or blinding (aka masking) if you are not already. Implementing these in your experiments reduces the impact of subconscious bias, making your results more reliable.
Feedback explains the implications of different experimental design decisions, allowing you to make an informed decision about your experiment.
If aspects of your design are unclear, EDA feedback will also prompt you to clarify this information. This ensures that your experimental design is represented clearly and that it reflects the experiments you are planning to perform.
The EDA provides statistical support when planning your study through the Analysis suggestion function. This includes suggesting appropriate statistical analysis methods for your experiment.
Analysis suggestions, and when each should be recommended, have been decided by a group of experts in experimental design and statistics.
Feedback can also highlight where there are issues (e.g. potential confounding), or where the experimental design is complex and consulting a statistician is advised.
We recommend you address feedback from the critique function before running the analysis suggestion as modifications you make in response to the critique might affect the analysis suggestion given.
Not all experiments require a sample size calculation. For example, if you are conducting a pilot study to test the feasibility of a new procedure, you will not need a formal power calculation to determine the sample size. Instead, you will need a rational justification for the number of animals based on how difficult you believe the new procedure to be, and how many animals you think you will need to be confident to decide whether this new procedure can be used in future studies or not.
If, however, you are going to draw conclusions from your data (i.e. use inferential statistics) the sample size should be determined by a formal method such as a power calculation or simulation.
Before attempting to determine your sample size, we recommend that you address feedback from the critique and run the analysis suggestion function. This ensures you have considered key aspects of your experiment, and helps you determine which sample size calculator is appropriate. This decision tree can also help you determine which sample size calculator to use.
Randomly allocating experimental units to groups helps reduce selection bias and makes your results more reliable. The EDA can create a randomisation sequence to help you randomly allocate experimental units to groups, this sequence takes into account any blocking factors or factors of interest that need to be randomised separately (e.g. sex or litter).
The randomisation sequence is sent directly to the person that will help you with randomisation and blinding. This means you can remain unaware of the group allocation throughout the experiment, further reducing bias.
Before creating a randomisation sequence, we recommend you perform critique on your experiment, run the analysis suggestion function and then determine the appropriate sample size. Knowing your sample size is essential enable a randomisation sequence to be created.
Not all experiments require experimental units to be randomly allocated to groups. For example, in a study phenotyping a mutant animal compared to a wild type you cannot allocate which animals are mutant and which are wild type (this is randomised by mendelian inheritance). You can, however, randomise other stages of the experiment, such as the order in which tests are done (e.g. behavioural tests) or measurements taken. You can also use block randomisation to control for litter or cage effects.
Communicating your experimental plans clearly is critical to facilitate review and improve reproducibility. To make this easier, the EDA can generate a pdf report, called the Experimental design report, containing the key experimental design information.
The experimental design report clearly lays out information requested by many funders and ethical review panels when assessing experimental plans.
In the report, experimental design information is displayed in tables, using a standardised format, making specific points easy to find for people reviewing several proposals. The report finishes with a visual overview in the form of an image of the experiment represented as a flow diagram.
You can use the experimental design report as an aide memoire when writing grants or protocols for ethical review to ensure all key information is included. Some funders and ethical review bodies allow you to add the experimental design report as an attachment.
The experimental design report can also be used in publications, to provide detail on the experimental plans.
If attachments such as pdfs are not allowed the read-only EDA diagram can be shared via a URL.
Communicating experimental plans thoroughly stimulates discussions with colleagues and collaborators and is crucial to facilitate review and improve reproducibility.
Experimental diagrams in the EDA give an explicit description of experimental plans. You can create a read-only version of your EDA diagram that can be shared with others via a URL, without the recipient needing an EDA account.
The read-only diagram allows recipients to click on a part of the diagram to see further detail about that element of the experiment. It also contains a table of key experimental design information.
Include the URL in grant proposals and applications for ethical review as well as publications to provide comprehensive experimental design information.
If URLs cannot be used in your applications a pdf, the experimental design report, can be created and added as an attachment. The experimental design report contains key information about experimental plans but does not have as much detail as the read-only EDA diagram.
The EDA is iteratively improved and updated, including the feedback provided by the EDA. To keep up-to-date on the latest updates to the EDA follow @NC3Rs on twitter or sign up to the NC3Rs newsletter.