The DSC Tg is defined as the heat capacity shift midpoint and the DMA Tg , as the average of the E" and tan delta peak temperatures measured at 1 rad/s. Both techniques are useful to measure Tg with up to +/- 9 degree temperature difference.
For more information,
Ferrillo, R. G., & Achorn, P. J. (1997). Comparison of thermal techniques for glass transition assignment. II. Commercial polymers. Journal of Applied Polymer Science, 64(1), 191-195.
The Tg measured with DSC and DMTA are not comparable quantities.
In DSC experiments the measurement of Tg should be done by cooling from the melt to the glassy state at given cooling rate (and this must be specified in any report). In DMTA the tests are influenced by both the cooling/heating rates and the test frequency. Changing the test frequency by a decade my give rise to Tg change of order of 4-5°C. The heating/cooling rates have similar effects. That is to say that if one uses very low testing frequency, f, and very low heating rate, H (namely , F=0.001Hz and H=0.1°C/min) the measured Tg may vary also 40-50°C in respect to faster heating and higher testing frequencies (namely, F=100Hz, H=10°C/min).
In all, the glass transition is a kinetics driven phenomenon . Thus is depends on the rates of testing (cooling/heating rates , test frequency and so on..)
Ways to delay the polymer chain movement with temperature will likely increase the Tg. The trick is to find the best possible way.
Both DSC/DMTA can be used to measure Tg. Personally I go for DMTA as it can monitor from the onset to the tan delta max which gives me a wider data set for comparison whilst analyzing different samples of the same class.