In RDA the dip averages are calculated by three methods
Poles, or normals, to dips are used for averaging and then the averaged pole is converted back to the averaged dip. The underlying reason for this is that a pole uniquely represents the dip plane, while the dip is really a line within the plane. Following is an example of why poles are used instead of dips. The average of a 5 degrees eastward dip with a 5 degrees westward dip should be horizontal. A vector mean of the dips yields a vertical dip, while a vector mean of the poles yields the correct, horizontal dip.
The different averaging methods have their strengths and weaknesses. Vector mean is a widely-used statistical technique for dip averaging. Vector median, although it is not as rigorous statistically, is generally superior to the mean and eigenvector methods for generic dips because it is less affected by "wild" values. When used in smoothing, vector median also seems to preserve the character better than either vector mean or the major eigenvector. The major eigenvector method is superior for averaging non-directional features such as fractures. For example, consider two fractures, one dipping 85 degrees East and the other dipping 85 degrees West. Although their dips are in the opposite direction, their dip planes differ by only 10 degrees. The eigenvector method treats these dips as if they are pointing roughly the same direction, while in the mean and median methods they will nearly cancel out. On the other hand, consider a normal and overturned bed dipping the same direction. These should not be treated as pointing in the same direction, but should cancel out. Thus, the vector mean or vector median should be used on bedding in preference to the major eigenvector method.