This semester I thought an undergraduate laboratory to last year engineering students. These will be entering the job market (or pursue graduate studies) in a few days. I was really surprised going over the lab reports how very basic things are omitted. Things like error analysis or displaying error bars on figures. How can one make any valid statement on the results or general behavior of data without them?
While the final format is not unique a few elements are quite common
- Your report will likely contain an introduction, review of the theory, method and materials with measurement set-up, results presented and discussed and a conclusion.
- If a calibration of an instrument is involved, shows the graph, give the mathematical relationship with the fit parameters and the quality of the fit (R2 at the very least). Do not forget the error bars.
- Make your figures readable! The experimental measurements should be represented by symbols while any fits should be represented by lines. Please avoid Excel. It can make acceptable figures if you work hard on them but specialized software (even GNUplot) are way better.
- Do not underestimate the power of well design tables to present a resume of multiple measurements. Going over 10 or more pages of lab report to be able to sum your results is not a good way to get extra points. Furthermore, it is likely that this exercise will lead you to make a better discussion of (or find obvious mistakes in) your data and analyses.
- To present (or unearth) trend in data, nothing beat graphical representations: do figures as often as you deem necessary and decide afterward to include them or not in the report. Again you might uncover interesting things in your data, which you have not thought of.
- If data are expected to be compared to a theoretical model, display the results of that model along with the data (see previous comment!). If the model can be fitted to the data, again give the fit parameters and the quality of the fit. If the fit parameters are related to physical values, they also need to be discussed.
- If data are to be compared to expected values or a theoretical model, your discussion should be quantitative ( not qualitative). Is the match perfect (within the error of x%)? If deviation are important, gives the deviation either in the form of absolute or relative differences or even replotting the data relative to the expected reference values or theoretical model. In any case, you are expected to make your discussion quantitative be default.
- Any deviation or unexpected behavior in the data must be discussed (even if you do not know the reason for a deviation)
While the above might seems a lot of work, it will in fact save you work when it is time to write your discussion, not mentioning that you should see an immediate improvement of your grades.