I have noticed that many good journal/conference papers (including some research books) have some non-smooth curve plotting, such as for packet delivery ratio (PDR), End-to-End (E2E) communication delays and packets reachability (RE).
In many cases, some sudden jerks in achieved values are noticed. The overall curve would show a meaningful result, such as fall of PDR with increase in network density, but it would also have few values in between going out of the fall trend. E.g. PDR for nodes = 400 would be higher for PDR for nodes = 350.
On the other hand, I have also noticed that many papers have reasonably smooth curves for the above mentioned metrics. What are factors that play a role in these non-smooth outputs? I personally could think of the randomness introduced in each simulation runs as being the major factor.
Thank you Daniel for the comprehensive answer. I ll surely keep your points in mind. I value your contribution.
Regards,
I'd like to emphasize that it is not always possible to make all your curves absolutely smooth, as was mentioned by Daniel. But even more important for a researcher is to know that it is not always essential, e.g. in the following cases:
1. The spikes can be the most important observation you get analyzing your results. In such a case, smoothing the spikes would be not a good idea and would provide the reader with wrong impression.
2. Furthermore, smoothing is not needed when the trend is obviously and clearly seen anyway.
- Badges
- Science topic