I am running Differential expression analysis on stringtie output using Ballgown.

My pData is :

head(phen_data_Early)

I extracted fpkm values using texpr, and have calculated manually the foldchange and log2(Fold chnage),the results are depicted below:

When I am running ballgown differential expression analysis using stattest, the foldchange values are different.

Result:

as we can see for transcript LOC_Os01g01010.1, the foldchange calculated using ration of fpkm values is 1.12 and fold change calculated using stattest is 0.802.

when I validated with deseq2 the foldchange value of LOC_Os01g01010.1 is 1.112 whic is almost close to 1.12(calculated using ratio of fpkm).

There was 99% correlation between log2FC calculated by ratio of fpkms and log2FC of deseq2. But only 50% correlation between stattest and deseq2.

Where did I go wrong? Please help me regarding this, It would be really helpful. Thank you.