Compression ratio

If your indexed data is reasonably uniform in size maybe you could calculate the average document size based on a sample set and use this to estimate the compression ratio?