Pick up your phone! It's vibrating!

JavaScript, 70 63 bytes

2 bytes saved thanks to Luke

a=>a.replace(/./g,a=>[['Rr','rr','-',' - ']['onp '.search(a)]])

Try it online!

Pyke, 22 20 bytes

cFh.o6.&\R*\-|l4)J" - 

Try it here!

c                      -  split(input, " ")
 Fh.o6.&\R*\-|l4)      -  for i in ^:
  h                    -        ^[0]
   .o                  -       ord(^)
     6.&               -      ^ & 6
        \R*            -     ^
           \-|         -    ^ or "-"
              l4       -   ^.title()
                 J" -  - " - ".join(^)

The crux of this answer is the transformation of ["long", "short", "pause"] into [4, 2, 0]. It gets the code point of the first letter of each word and ANDs it with 6. By lucky coincidence it transforms to the values we're looking for. (I searched through quite a few other longer solutions before finding this one). Once that's done, we can further transform that list of ints into ["RRRR", "RR", ""] by multiplying our int by "R" which then turns into ["RRRR", "RR", "-"] and finally title casing it to get ["Rrrr", "Rr", "-"]. We then join the resulting list by " - "

Haskell, 71 66 59 bytes

g 'o'="Rr"
g 'n'="rr"
g 'p'="-"
g ' '=" - "
g _=""
f=(g=<<)

Try it online!

Oh right, =<< is concatMap.

Takes advantage of the fact that "long" and "short" both have the letter o.