Recent studies have shown that microRNA (miRNA) inhibitory activity can be quantified by examining their target mRNA expression levels. The accumulated evidence of differential miRNA activities between cancer subtypes necessitates the systematical comparison of miRNA expressions and activities. In this study, we integrated 8 mRNA microarray datasets to infer and compare the miRNA activities between prostate cancers (PCs) and normal tissues (NTs). Gene expression analyses show that miRNA activity is stronger in PCs. This conclusion is consolidated by target protein expression. We simultaneously collected 6 independent miRNA expression datasets, where great inconsistency is present in the expression difference between PCs and NTs. The overall correlation between miRNA activity and expression is very weak. However, meta-analysis demonstrated that the expressions of 114 individual miRNAs agree with their activities. Additionally, we detected two other factors associated with higher miRNA activity in PCs. One is deregulation of some key miRNA-repression related genes, such as the over-expression of Dicer, TRBP and Ago2, and the under-expression of IRP1 in PCs. The other is that miRNA-mRNA binding site efficacy has significant positive correlation with miRNA activity, whereas no correlation with miRNA expression. Furthermore, miRNA activity is more reproducible than miRNA expression across different datasets, which allows miRNA activity to be a good feature for the classification of cancer subtypes. We expect our analysis can improve the methods for inferring miRNA activity and further, provide some clues to the role of miRNA in tumorigenesis.