Abstract　 　　
　 We have carried out stacking analyses at radio (VLA, GMRT), sub-millimetre (LABOCA, SPIRE), and far-IR (PACS) wavelenghts for large samples of z = 3, 4, 5, and 6 Lyman Break Galaxies (LBGs). The UV/optical/near-IR spectral energy distributions (SEDs) of these (rest-frame) UV-selected galaxies have been modelled in detail, resulting in robust estimates of their un-obscured star formation rates (SFRs), stellar masses, and dust-extinction. The latter can be translated into a model-predicted IR luminosity (assuming that all the UV photons absorbed by the dust are re-emitted in the IR), which in turn can be compared our direct constraints on the IR luminosity obtained from our radio/submm/far-IR stacks. We find a remarkable good agreement between these two distinctly different ways of estimating the IR luminosity. The UV/optical/near-IR models depend on the assumed star formation history (three different scenarios are assumed: a declining, a constant, and a rising star formation history) - and we have used the above comparison to determine the most likely star formation history. We also discuss our results in the context of the IRX-beta relation (Meurer+99). Final disclaimer: this is work in progress, and critique and/or input will be much appreciated.