Personally, I would want to see a "normalized" COMPARISON by volume size.
That means "decomposing" each of the bars into partial bars which if added together would "perfectly reconstruct" the "normal" original total volume bar. This method would show the effect of large versus small size trades in each bar with zero lag time of the comparison, and without error caused by fluctuation of overall trade sizes distorting the view of the relationship of largest versus smallest trades regardless of how big or small the average trade size is.
For example, you could make one set of bars from "VolumeA completed", and another from "VolumeB completed".
To make the bars comparable (normalized) , make VolumeA + VolumeB be total "Volume" to complete the bars, but VolumeA be the smallest blocksize HALF of the total volume, and VolumeB the largest blocksize HALF. Accumulate the price "deltas" ATTRIBUTABLE to each equal "tile" of the volume in each volume bar to generate the open-to close changes, save the excursions from open-to high and open to low attributable to their own volume tile, and you will get two synchronized charts with different shapes representing the activity of the two "tiles" of volume size. Both will add up to give the total-volume bars' open-to-close price change exactly, while the tick movements belonging with their own tile of the trade sizes will be allocated to their own special bars or candlesticks. Trade both charts, and call it trade size diversification. Or just see if the differences make a useful indicator.
Since I do not use Ninja, I hope someone can try this soon and post the result. I'm curious what the result would show!