Let's consider a data set in an $m\times n$ matrix $A$, where $n$ is arbitrarily large. For the sake of discussion, let $m=2$ and only think about any one column of $A$. The rows represent different methods for measuring a single property of $n$ objects. The data obtained from the top method varies much more than the data obtained from the bottom method. Let's express this variance as follows. Let $M_1\gg M_2$; the data $a_{1,j},a_{2,j}\in A$ are uniformly distributed in their respective ranges $0\leq a_{1,j}\leq M_1$ and $0\leq a_{2,j}\leq M_2$ (i.e. the stuff in the top row can be much larger than the stuff in the bottom row). $M_1$ and $M_2$ are unknown.
I want to combine $a_{1,j}$ and $a_{2,j}$ into some kind of average. It seems like a bad idea to take the arithmetic mean, as $a_{1,j}$ would be given more weight given its larger range (variance). What is a good way of combining the data in a given column into a single number?
I am not well-versed in statistics nor maths in general. Please explain like I'm 5 years old.