This needs a thorough analysis by carrying out assessments in a large number of patients to prepare a normative database. To the best of my knowledge, there exists no such database as such.

The data base shall be then compared for age and gender specific differences in mean values by using ANOVA and Student "t"-test. If we come across differences in different age groups or between two genders (in overall as well as different age groups) then we shall be requiring a different cut-off value (normative range defined by 95% confidence intervals) in different age and gender groups.

I have recently prepared three gender and age specific normative data bases for kidney size, ear morphology and carotid artery morphology using the same criteria and it really helps to understand the relationships.

The expected outcomes in your study are - females having a higher body fat % as compared to males, with increasing age - the normative range for % body fat shall increase.

Can you use normal values for young healthy, or old healthy people and compare to that?

I would think that any "cut off values" would be dependant on their purpose. Are you looking for cut off values for defining health risks associated with obesity? Muscle wasting in a certain patient population? Health or performance of competitive athletes? Each population has it's own set of variables which needs to be considered that is specific for that population. And of course, hydration and anything that effects it will impact measured bioelectrical impedence data. The first step has to be to define your target population and identify the variables likely to influence their outcome data.