I am using wgcna to create Coexpression Networks for different conditions. My goal is to do differential network analysis on these networks. Starting sample size is different for each network; however the genes are kept the same for each condition. Since I am using wgcna and Pearson's correlation to calculate the weight of the genes depends on the sample size. Since I am using the weights to calculate different centrality measures; the difference in these measures I am seeing reflects the sample size and not the actual difference in centrality. Even if I calculate intra-modular level- the module selection- still depends on the weights. Therefore- is there a way to decrease the effect of sample size ? Should I just have to keep the same sample size for all networks to be able to really see the difference in centrality measures due to condition.