The first suggestion in latkin's answer seems good, although I would suggest the less long-winded way below.
PS c:\temp> $global:test="one"
PS c:\temp> $test
one
PS c:\temp> function changet() {$global:test="two"}
PS c:\temp> changet
PS c:\temp> $test
two
His second suggestion however about being bad programming practice, is fair enough in a simple computation like this one, but what if you want to return a more complicated output from your variable? For example, what if you wanted the function to return an array or an object? That's where, for me, PowerShell functions seem to fail woefully. Meaning you have no choice other than to pass it back from the function using a global variable. For example:
PS c:\temp> function changet([byte]$a,[byte]$b,[byte]$c) {$global:test=@(($a+$b),$c,($a+$c))}
PS c:\temp> changet 1 2 3
PS c:\temp> $test
3
3
4
PS C:\nb> $test[2]
4
I know this might feel like a bit of a digression, but I feel in order to answer the original question we need to establish whether global variables are bad programming practice and whether, in more complex functions, there is a better way. (If there is one I'd be interested to here it.)