Ok Tadahiko-san,
It looks like the model is producing a gain of 0.79dB instead of 0dB at Vg = -1V midrange. So that is not the typical +/-0.4dB gain error in the data sheet, but within the worst case 25C of +/-2.2dB. I tested 0V and -2V and in each case just a little less than the +/-40dB gain expected - I think it is working, the absolute can acccuracy for this part is not that good - if you need better, look at the VCA821
Also you need to move the source to be in series with the 20ohm if you want bias current cancellation,
