Even a big underground test like the 5MT Cannikin does not produce a huge earthquake (I believe that was a very localized reading of 6.5), it is the signature of the device that is very recognizable. An earthquake is sort of like shaking something, whereas a detonation shows up on seismographs more like a very sharp blow from a hammer. No lead up, a sharp peak that "rings" down to normal very fast.
I know some seismic guys that pinpointed the Russian sub Kursk's explosion underwater, which was non-nuclear, shortly after it happened.
For an extremely small device, it would almost be easier to hide a test above ground than under it.
Independant of the type of "shaking", if the equipment was not in place to detect a 2.7, then can it be discounted based on seismic evidence?