Lincom/test using estimates from two independent samples?

Hi,

I would like to test if the unconditional mean of outcome Y from one sample is equal to the linear combination of a regression estimate (i.e., partial effect of X on Y) in a different sample. Below is a silly example, but it should demonstrate the problem.

Suppose I know the proportion of females aged 20-46 with high blood pressure from sample A:
*** Sample A's unconditional mean of outcome Y
webuse nhanes2, clear
keep if female==1 & inrange(age,20,46)
mean highbp [aw=finalwgt]
est sto A

...and I know the effect of race (black) on having high blood pressure in sample B
*** Partial effect of X on Y from Sample B
webuse nmihs, clear
reg highbp race age [aw=finwgt]
est sto B

I would like to find a linear combination using estimates from the two samples to account for their respective variances
*** Adjust sample A's unconditional mean using partial effect estimate from sample B
suest A B
lincom [A_mean]_cons - [B_mean]race*2

Clearly this doesn't work because Stata wants to call the data to construct the "stacked" variance matrix. Another problem I have is that one of the datasets is on a restricted server. I might be able to import the unrestricted data, but I want to first ask if there was a way to construct a test (Wald?) somehow by hand, simply using the variance matrices.

Thanks