Data compression is well known as a transformation of data traffic that is able to greatly increase the data rate in a network, and as the cost of processing power decreases compression is becoming more and more widely used for this purpose. In this paper, we look at the statistical properties of this transformation, studying changes in the probabilistic characteristics of packet sizes caused by compressing packets transmitted under three different network protocols (TCP, IP, and Ethernet) and using three different compression algorithms. The analysis is performed on real data traffic recorded in a local area network environment. Our aim is to provide network designers with a model of compressed data traffic that could be useful in modelling and performance evaluation of telecommunication networks.