Linear Interpolation

my data is like this:


header | data | key
-------------------
500 | 3.2 | 10
500 | 3.4 | 20
500 | 3.6 | 25
500 | 3.7 | 40
501 | 4.1 | 10
501 | 4.2 | 15
501 | 4.4 | 30
501 | 4.6 | 35


and what I want to do is find the median of "data", but keyed off of "key", so if my desired median is 30, I want to take the two records (data, key) nearest to key = 30, and get the average of "data".
...and do this within each "header" value.

actually, to be precise, I want the linear interpolation, so for header = 500, I want to get the (data, key) pairs of (3.6, 25) and (3.7, 40) and return the interpolated "data" value of 3.6333 (as done here (http://en.wikipedia.org/wiki/Linear_interpolation))

so for the above example the query would produce:


header | interp
-----------------
500 | 3.633
501 | 4.4


possible, or am I crazy?