NavList:

There are two basic sources of uncertainty in dip: your ability to estimate your height of eye accurately, and anomalous refraction. These are directly apparent in the dip equation (apologies for working in feet):
  dip = 0.915 · (1.06 · sqrt(HoE in feet))
usually quoted as
  dip = 0.97 · sqrt(HoE in feet).

The factor of 0.915 is actually variable and results from the refraction of nearly horizontal rays of light from the observer to the horizon. Quite ordinary changes in refraction close to shore can change this factor to 0.89 or 0.94. Some authorities claim that this variability can be reduced by shooting from a higher height of eye. I have never seen any real tests of this claim, but it's possible.

The dependence of the dip on height of eye is obvious, but the sensitivity to uncertainty can be analyzed in a few different ways. For example, if I assume I have a "one foot" uncertainty in height of eye at every possible height of eye, then clearly the sensitivity decreases when you're higher up. So stay high... But suppose I assume I have a 5% uncertainty in height of eye at every possible height, which is probably realistic. That translates to a 2.5% uncertainty in dip which would be proportionately larger in dip error in minutes of arc when you're higher up. So stay low... In short, there is uncertainty in determining the uncertainty!

On top of this mathematical sensitivity, there are observational issues. From a higher observing location, the horizon is further away and the wave tops which compose it show less irregularity. You're also less likely to be fooled by an intervening ocean swell in the distance.

Frank Reed
PS: You asked about height of eye on typical sailing yachts. There's no guarantee, but figure 3 meters (10 feet) or higher.