Tuesday, 23 March 2010

How to Present Research Data: A Lesson!

I guess I am not only the prize sap, and may not only be this time. I think I may be ignorant in the past too. I was doing experiment on real testbed last summer and was logging all the real time data for later analysis. I predefined my mind that what I need to discover from this data, I was looking for time delay of the traffic and throughput of all data. I plotted all graphs and I thought I am done with the experiments. Though, the throughput I got was not good, but I was able to justify the reasons and there were lots of spikes in the graphs . I knew "why these spikes are?" due to my one set of topology. I repeated the experiment 100 times but because setting were same, so didn't get the significant difference. I closed the results in my research report and moved on...

Now the things after a year:

1. I didn't mention anywhere in my report that why these spikes were, though I know it..but I should have mentioned it explicitly for the reader. Isn't?
2. If I knew that spikes are due to one set of topology, I should have tried to normalize results by making different topologies. But I was too lazy to do that, so I convinced my self, that if somebody enquire about it, why these spikes, I'll clearly say that due to topology. But I didn't bother to think that if somebody asks , did you try with other topologies or Why you chose this particular topology? Is it standard? So, now as a result, I learnt that either I had have selected standard topology or provided normalized results. Should try to give answers of all the questions in written form in the report before anybody asks you.
3. When I was analyzing results(I mean drawing graphs etc.), I found sink has more data from nearest node than distant node, but then I thought, so what?
Its obvious that near nodes traffic reaches earlier and suffers with less error rates. But Again, I didn't write my this observation in the report. If I would have drawn a graph to show in what proportion this occur and more reasons..it would have been a publishable work!
So, I hope I'll learn from my own mistakes, but wish you shouldn't repeat the same :).

No comments:

Post a Comment