The answer to your question is that statistical analysis is used to control measurement error in all phases of chemical research. This gives the chemist much better information than the old analytical rule I was taught: do it three times, you may discard one observation if it appears off and report the mean.

What statistical method, eg ANOVA, depends on the particular research question, in this case the purpose of the experiment which I don't know. It's generally better to control variability than not. Best wishes, David Booth

Dr.David,

Thanks for responding, the observation got where the conc. Of a solute decrease can be calculated directly by subtracting the final conc. Of solute by the initial conc.present in a solvent ,so is there still any need to convert this data to check hypothesis?

If you don't have multiple measurements ok. However if you have more measurements I would take variability into account. see the attached reference in the z-library. Best wishes, David Booth

Dear Anika, many thanks for asking this interesting technical question. As we are synthetic inorganic chemists, please don't expect an expert answer from my side. Better trust the advice of a proven specialist like David Eugene Booth. In our field of research the presentation of data is rather dfferent (e.g. elemental analysis, spectroscopic data, X-ray diffraction data) and there is no need for statistical analysis. I would just like to suggest to you the following potentially useful article which migth help you in your analysis:

Why we need to report more than 'Data were Analyzed by t-tests or ANOVA'

Article Meta-Research: Why we need to report more than 'Data were An...

This paper has been posted by the authors as pubic full text on RG. Thus you can freely download it as pdf file.

Good luck with your research and best wishes, Frank Edelmann

Thanks Dr. Frank

I will go through it.

There are mainly two reasons:

1) to see how likely your conclusion (i.e. "treatment reduces the OD") can be fooled by the data and

2) to say how precisely you could estimate the treatment effect.

Frank T. Edelmann : if I remember correctly, data in X-Ray diffraction are given with an uncertainty, for instance the atomic coordinates or the unit cell parameters… So I would not say that statistical analysis is not needed, instead that it is used differently, and may be more transparently for the user…

Same thing apply I guess for elementary analysis and spectroscopic data, at least for the rounding of the data at a correct number of figures… In a way, a kind of statistical analysis is already performed "internally", to calibrate the apparatus and so on.

Just my point of view ;)

Dear Emmanuel Curis many thanks for your kind response and comment. Yes, that's certainly a good point. In our field of research (synthetic chemistry) either the software of the machines (e.g. X-ray diffractometer) or the specialist co-workers and technicians who run the machines do the statistical analyses. This should not be forgotten.

Just to add a little to the discussion. Calibration is a separate statistical method related to regression but different from it. The x and the y are reversed causing all the Statisical procedures to be different. BTW practicing chemists need to be aware of that so spectroscopic instruments, etc are calibrated correctly. The attached screenshot search gives more details.

Your friendly neighborhood physical chemist/statistician, David Booth. BTW R Has very good calibration procedures.

Dear Dr. Frank,

Yes I have same opinion as we always perform experiment in triplets and take mean then apply the mathematical formulas .

But statistics is different.

The generic response is that all measurements have some degree of uncertainty and by repeating the measurements, independently, several times, that uncertainty is revealed.

Statistical analysis removes biases that the researcher can have in his assessment of the results.

Surveys and research designs can lead to mistakes if the data is not analyzed correctly, helping the researcher avoid pitfalls such as stating that a variable has a causality relation to the response when there is only a correlation.