If a scan is performed and the final scan has measurements that are missing, most likely our measurement prediction algorithm hid those measurements. Those measurements were hidden because a measurement value was out of a range or
Given height weight gender and an age range, we have an algorithm that calculates the predicted measurement, similar to the way the body fat algorithm works. We had an input dataset of around 20,000 scans, and linear regression was run for each measurement for each age group (18-25, 26-35, etc) to derive a predicted measurement. We then take a range on either side of the measurement and compare the incoming new measurement with the predicted measurement range. If the new measurement falls outside of the range, we flag it as bad.
The reason we added this feature was to be able to detect when someone was wearing loose shorts or if their hair was down. For example, if someone forgot to put their hair up in a bun, we would see a bigger than expected neck measurement. This measurement would be outside of a certain bounds. So we would then hide the measurement and give a predicted measurement based on the criteria mentioned above. This is so data isn't inaccurate when someone forgets to out their hair in a bun, you’ll instead have a measurement closer to the true value.