i hava the RNAseq coutns data and fpkm data from TCGA, i want to know the fold change of tumor tissue RNAseq expression contrast of normal tissue RNAseq expression ,counts data is the RNA reads number, fpkm data is similar to real RNA expression. i found that the logFC results from DESeq2 and edgeR analyse using counts data  are different from that i use mean tumor fpkm data devide by normal fpkm data, i found that fold change  in fpkm is sifferent from the DESeq2 and edgeR fold change results . which fold cahnge  is better to choose ?

Your own simple logFC computation is not likely to be as good as the more sophisticated computation done by edgeR (or DESeq2).

LogFCs are computed by negative binomial generalized linear models.