Tri-quadratic LINEST...possible?

I know a lot of us try really hard to use as few cells as possible. When I do regressions using LINEST(), rather than create the "known_x" matrix directly in the LINEST() function like you have been doing, I prefer to use a block of helper cells to create the known_x matrix. For a regression like this, I would have

regression column
1 -- 2 -- 3 -- 4 -- 5 -- 6
=OAT -- =OAT^2 -- =WB -- =WB^2 -- =OD*WB -- =SH
copied down.

Then, my known_x argument is simply a reference to this block of cells.

If you really need to have this in a single cell without the block of helper cells, someone else will have to help you put that together.

Which is a good part of why I like to use a helper range for this.

With a helper range, adding terms to make this a tri quadratic (whatever that looks like exactly y=ax^6+bx^3+c maybe?) is simply a matter of editing/adding columns to the helper range. When you try to do it in a single cell, you have to figure out how to formulate the function to add those columns inside of a single cell. I'm sure this can be done in a single cell. I just never bothered to develop the skill of building complex single cell formulas like this.

What's the form of the resulting formula?

y = a*? + b*? + c*? + ...?

Once you decide,

=linest(y1:y100, choose({1,2,3,4,...}, x1:100, z1:z100, x1:100*z1:z100, x1:100*w1:w100, ...), , TRUE)

... for y = ax + bz + cxz + dxw + ...