The indirect learning architecture is one of the most common setups for the identification of digital predistorters which are used to linearize radio frequency power amplifiers. To apply the indirect learning architecture in a real transmitter, the unknown delay and gain between the transmitted and the received signal must be corrected. In the present paper we present a survey of delay and gain correction methods and discuss basic objectives for selecting the delay and gain of the linearized system. We demonstrate the equivalence of gain correction before and after identification and conclude with a method for performance comparison at constant output power.