Depending on how random your sample has to be, I'd suggest getting some random numbers and then adding some of your own to get the desired median and mean. There's a positive skew, so you may want to ...