If you follow the IEC 61400-23 recommendations (which are for power-curve testing) then a simplistic ratio approach is recommended. However, this standard is intended for normalising datasets at 'near sea level' conditions - not for detailed energy estimates. In reality, a stall-controlled turbine will be tuned for site, with the pitch setting feathered for lower density so that rated power is always hit.
For example, if we take a 2000m site (p=0.99 kg/m3) and correct it using IEC measurements we see a large decrease in energy (around 25%) - whereas the actual power curve will be around 10-15% decreased after adjusting the pitch settings. In this case, the pitch settings were feathered by around 2 degrees to achieve this power curve.
Therefore, the site-specific power curve should always be used for energy estimates, and this should be sourced from the manufacturer. For sites at higher altitude (I would say above 1000m), then tested power curves should be insisted on as the aerodynamics of the turbine will change (particuarly for stall-controlled turbines) which invalidated the original testing. I'll go into more detail on stall-controlled turbine aerodynamics in a future post.