Wednesday, August 13, 2008
Plotting Data
I spend a lot of my time extracting data from different graphs, manipulating the raw data in various ways and generally plotting and pimping graphs so that they are clear and easy to understand. Personally I prefer my data presented in as close to raw format as possible with maybe just the odd modification to correct baselines and remove noise so that the plots are simple, clear and easy to interpret. While the data can be manipulated to give a straight line I much prefer to see the data in the original plot as it's much easier to spot deviations or anything strange which might be going on. I understand that before the days of computers for everything, rearranging data to give a straight line much simplifies things, but with programmes such as SigmaPlot, OriginPro, Igor, and Prism to name but a few pretty much any data can be fitted to quite complicated equations so there is no reason to do double reciprocal plots or whatever.
When I read a current paper which determines constants using double reciprocal plots or other methods rather than fitting to the original plots I can't help but be rather sceptical. Either they really know nothing about the subject and don't realise the potential errors of analysing data this way (which doesn't instill confidence in the rest of the paper) or they know their data is a bit shonky and this is the only way to get a fit out it (which again makes me sceptical and wonder what they are not saying). I recently saw that some of my data doesn't fit to the commonly accepted model and rather than forcing my data to fit by using different plots or removing data points, I just submitted it without any fit and then commented about why it doesn't seem to fit and deviates from general accepted principles - I just hope the reviewers think like me!
When I read a current paper which determines constants using double reciprocal plots or other methods rather than fitting to the original plots I can't help but be rather sceptical. Either they really know nothing about the subject and don't realise the potential errors of analysing data this way (which doesn't instill confidence in the rest of the paper) or they know their data is a bit shonky and this is the only way to get a fit out it (which again makes me sceptical and wonder what they are not saying). I recently saw that some of my data doesn't fit to the commonly accepted model and rather than forcing my data to fit by using different plots or removing data points, I just submitted it without any fit and then commented about why it doesn't seem to fit and deviates from general accepted principles - I just hope the reviewers think like me!
Subscribe to Posts [Atom]