Silicon nitride exhibits fatigue based on cyclic crack propagation which is critical for components under repeated loading. Lifetime predictions for such components are usually based on power law formulations and are most sensitive to the crack growth exponent n. Various statistical procedures exist to determine the parameters from cyclic tests of un-cracked smooth samples. In this work, an analysis is presented for silicon nitride (SL200) lifetime data with a focus on the relation between load ratio and the exponent n. It is found that n increases with the load ratio which has also been observed for macroscopic cracks in silicon nitride earlier. A high degree of uncertainty is associated with the crack growth exponent due to the sparse lifetime database. A pooling strategy is presented which increases the sample size of the underlying lifetime distribution and successfully decreases the scatter in the crack propagation curves and the uncertainty in the crack propagation exponent and, thus, in the lifetime prediction.