## DISCOVERY PERFORMANCEâ€‹ï»¿

AutoDiscovery is designed to find out relevant relationships within your consolidated experiment data. To do that, it performs a variety of statistical and computational techniques at different levels of detail.

The discovery performance of AutoDiscovery depends on the following parameters.

Discovery Performance

Depth Of Analysis (max)

1

Ratio Variables

NO

Parallelized Calculation

NO

3

YES

NO

5

YES

YES

Depth Of Analysisï»¿

Relationships are evaluated in every subset of your consolidated data which are obtained by filtering the samples using the different values of the qualitative variables detected.

E.g. in a classical animal behavior assay, possible subsets would be the animals of each particular group, or the trials recorded with each of the substances injected.

AutoDiscovery can combine the qualitative variables found in the consolidated data in groups of up to 5 variables so that the subsets obtained are more and more refined (e.g. animals in the control group AND injected with the substance X).

The Depth Of Analysis is exactly the maximum number of variables that AutoDiscovery will combine to obtain the subsets.

If your experiment was designed to work mainly with quantitative (numerical) information or if the number of samples captured is relatively small (less than 20), an adequate Depth Of Analysis would be 1-2.

On the other hand, if your experiment was designed to work mainly with qualitative (categorical) information or the number of samples captured is large (more than 100), a Depth of Analysis of 3-5 would be the ideal.

Ratio Variablesï»¿

Many times, relevant discoveries are not found on relationships between two single quantitative variables but between a variable and a combination of others.

Ratio variables are automatically generated by AutoDiscovery when the discovery process starts by combining every pair (A, B) of quantitative variables and calculating the quotient between them (A/B).

The resulting variable is automatically labelled as "Ratio A by B" and is internally used in the discovery process as other quantitative variable although it is not shown in the consolidated data table.

Ratio variables dramatically increase the probability of identifying a relevant discovery so make sure to select the adequate version of AutoDiscovery to maximize your success!.

Parallelized Calculation

This computational technique allows AutoDiscovery to share the calculation workload among different computers.

That means that if a process took 60 min in your computer, parallelizing the same process among 3 computers will make it to take only 20 min (aprox.).

Parallelized calculation is a special service provided exclusively for Research Centers which consolidate large amounts of information coming from different labs and require minimizing the processing time drastically.