The US Military (and others I am quite certain) extensively study casualties, including those in WWII. Military planners have long used the number of troops engaged and duration of combat as the primary predictors of casualties. In these circumstances, the number of troops involved derives from the planned correlation of forces (i.e., enemy strength, terrain, etc. leading to a determination of required friendly forces) and expected duration of combat (i.e., planned time to complete the mission). As a result, casualty rates are generally presented/analyzed as casualties/1,000 combatants/day.
I think there are simply too many factors involved to simply predict the duration of fighting in the home islands from casualties, number of opposing personnel, days of battle in the island-hopping campaign and time since Pearl Harbor. It considers nothing of the land area and type of terrain in the battle ground. A lot of the Pacific battles were fought in small areas with high troop densities in bunkers. Once a beach head is established the battle takes on a different character.
The following table comes from a US Navy publication concerning casualty rates in the Pacific:
The Navy's general conclusion is that ground casualty rates were higher during the short duration operations early in the war. Tarawa being a prime example. For these amphibious actions, casualty rates are highest on the first day and fall rapidly thereafter. The Navy conclusion was that the Marianas campaign represented the type of battle that would be fought in the home islands. However, once the landing succeeds, the casualty rate dropped significantly
The only exception to this was Okinawa where the Japanese chose not to oppose the landing. I would also point out that casualties quoted in Wikipedia must include sea, air and ground casualties since the computed rate (0.83) is lower that the USMC ground combat loss rate (which was lower than the US Army loss rate - which I do not have at my finger tips).
But these are only planning numbers for staff use, to help determine logistics, medical and replacement requirements. Manpower requirements for the next year... It certainly gets factored up by the size of the enemy force (I don't have that equation, and don't believe there is a single accepted model).
Duration isn't always a clear factor - the Red Army suffered over 350,000 casualties in the last 23 days of the war and their assault on Berlin and central Europe.
In any event, do the casualty rates indicate that the IJA is "performing better" than earlier in the war? No.
Are they worse? Impossible to conclude, but most believe that the early action reflects US inexperience more than Japanese skill; and that the general trend of decreasing US casualty rates in the Pacific was probably do to US experience and fire power increasing rather than any change in IJA combat capability.
In early 1945, the Army Strategic Planning Staff in Washington operated from what they called the "Saipan Ratio" (have not seen the equation/model). This led to a lot of argument with MacArthur and casualty estimates so high as to defy imagination. In fact, based on the opinion of "senior officers" the "Saipan ratio" was rejected for planning purposes because casualties could not be so "huge as to make the task insurmountable".
When you examine the amphibious operations involving the US Army in the PTO, the average casualty rate is reported to be 7.45. This is a different dataset than either the table above or the Wikipedia-derived numbers. The comparable number from the ETO was 2.16.
Obviously MacArthur's staff argued for something closer to the ETO casualty rates (because he wanted to invade), while the Navy argued for something like the Saipan casualty rates, which they believed were more realistic (which in turn means that an invasion will be terribly costly, ergo blockade).
I have seen references to "current computer models" that project casualty rates to estimate medical sustainment requirements. The factors included in the model include:
Number and type of troops (CS and CSS troops have lower casualty rates)
Projected duration of the conflict
Opponent quality/capability
Intensity of combat
Terrain
Climate
Morale/cohesion factor estimate
My understanding of the program (FORECAS) is that it uses simulation techniques to create a range of predictions. None are estimated using simple linear correlations.
Good discussion, but I believe the approach is too simplistic.
Best regards,