Assuming you don't use a buck switching preregulator, its even worse. The current waveform in the secondary has a high crest factor due to the bridge only passing current near the peaks of the voltage waveform. A simple LTSPICE simulation shows that anything under 24V RMS 80VA is going to be unsatisfactory. Also that makes no allowance for low line conditions - if you need the PSU to be able to produce its rated output with the mains input 10% low, you'd better use a 26V RMS 100VA transformer with double the reservoir capacitance.
This simplistic model does assume that the transformer regulation can be modelled as a resistive loss transferred to the secondary, but that should be good enough to size the transformer appropriately.
.model 1 D(Ron=1u Roff=1G Vfwd=0) ;Ideal diode
.model 2 ako:MURS320 ;Si 3A diode
;.step param D list 1 2
.param D 2
defaults the diode model used to MURS320, but if you remove the ; commenting out the .step, it runs the sim twice, first with an ideal diode and then with the MURS320.so you can see the typical diode drop for the bridge. Unfortunately, LTSPICE cant average or calculate the RMS (ctrl-click a waveform legend) for .stepped simulations, even if only a single step is selected, so if you want to do stuff like check the RMS current in the secondary V1, you need the .step commented out.
Comment out the .ic and change the .tran simulation command to .tran 0.2 to see the initial inrush current to get an idea if you are going to need a soft-start circuit. If you need better resolution on the current peaks, set the max timestep to 20us., at the expense of sim running time, but that's pretty much gilding a turd as the transformer model isn't good enough for accurate numerical modelling, which will also be highly dependent on the actual diode used. Worst case, with a 10000uF cap and a 26V 100VA transformer with the line 10% high, its going to be over 50A, so you are either going to need a very beefy bridge or a soft start circuit.