Improvements to Normative Modeling; Normative Models Account for Sex and Handedness
In Dexterit-E Explorer 1.4.0 (and Dexterit-E 3.7.0), we have extensively reworked our normative modeling process resulting in a significantly more robust comparison to our modeled healthy behaviour. In Dexterit-E 3.6.1, we stopped accounting for sex and handedness effects to help reduce modeling errors that were introduced by splitting data sets into small sample sizes. As promised in September 2016, we have now developed our long-term solution, which avoids the problems of modeling errors introduced by data set splitting.
The new modeling process allows us to re-introduce sex and handedness effects in our normative models while also reducing modeling errors over previous versions. The impact of these changes in terms of “what to expect for a given Z-score” was quantified. As an example, for Z-scores of ~1.64, the mean of the standard deviation of Z-score changes from analysis version 3.6 to 3.7 across all parameters was 0.31 (i.e. typically the change will be less than 0.31).
In addition to a new modeling process, we have significantly increased the number of subjects in several of our normative databases and added modeling for the new KINARM Exoskeleton Lab released in 2016.
These methods and our verification process are documented in detail in both the Dexterit-E Explorer User Guide, and our new KINARM Standard Tests Summary. This latter guide is available publicly on our website. To ensure correct representation of our methods, we encourage you to cite the KINARM Standard Tests Summary in your manuscripts as: