It depends where the marker is located and if it's a cyclic movement: if the marker is part of a (more or less) rigid body, for example the shank, and you have other markers on that rigid body that are still visible, you can generate a virtual marker based on the markers that you can still see (because you know the position of the missing marker with respect to the visible one). If it's a cyclic movement, e.g. walking, you can extrapolate from previous cycles. If it's neither of the two, you could extrapolate for example in Matlab after exporting the marker data, but of course you wouldn't know if this extrapolation is correct.

I assume you've captured your data on Vicon Nexus, if that is the case, it would be more convenient to use the gap filling functions on the Nexus software itself. There are currently 5 gap filling options in Vicon, Spline Fill, Pattern Fill, Rigid Body Fill, Kinematic Fill and Cyclic Fill. Spline and pattern fill is often used when the gaps are small (less than 5 frames). If the gap is around 5 - 15, pattern fill can generally be used. Rigid body fill can only be used if the marker is attached to the assumed "rigid segment" with 3 other markers attached to the same segment. However, you need to take extra care with the assumptions you are making when using rigid body fill, taking into account the skin movement artefact as well. 

I'm not so sure about the gap filling options in AnyBody software as I have never used it before. However, extrapolation of missing markers can be easily achieved in Matlab using the recommended filling algorithm on the Vicon website, as suggested by Andreas Kramer.

http://www.vicon.com/faqs/software/what-gap-filling-algorithms-are-used-nexus-2