PerspectivesAre you interested in submitting a Perspective Article? Be sure to read The Science Advisory Board's Editorial Guides for Perspective Articles. Click here. Intuitive statistics for easy data analysis: A Review by Gabriel Dorado1, Pilar Hernández2, María del Pilar Dorado3 1Author for correspondence, Dep. Bioquímica y Biología Molecular, Campus Rabanales C6-1-E17, Universidad de Córdoba, 14071 Córdoba (Spain), eMail: bb1dopeg@uco.es; 2Dep. Agronomía y Mejora Genética Vegetal, Instituto de Agricultura Sostenible (IAS-CSIC), Alameda del Obispo s/n, 14080 Córdoba; 3Dep. Química Física y Termodinámica Aplicada, E.P.S., C/. María Virgen y Madre s/n (esquina Menéndez Pidal), 14071 Córdoba The diversification and specialization of current science and technology represents an strategic advantage, but may also pose a significant handicap for research, discovery and advancement. An example of that can be found in the life science disciplines, which some times require also skilled mathematical knowledge to analyze the data generated on the research experiments. Life scientists may master their particular and specialized research area, but unfortunately, they do not always master mathematics in general and statistics in particular. Thus, such knowledge divide may represent a strong limiting factor for many researchers. Fortunately, there are different approaches to overcome such a situation. One of them is to hire or -better yet- collaborate with a statistician, to do the math work on the research results. Such an approach may be quite productive and useful, enhancing the transversality and multidisciplinary nature of current research. Yet, there is also another alternative, which is not mutually exclusive with the previous one. Namely, to exploit the possibilities offered by the new information and communications technologies (ICT). Among them are the computers in general, and in particular, specific applications created to make statistics easier "for the rest of us". Which is the best statistical package? The answer to such a question depends on both what the users want to do and their statistical background and expertise. There are a bunch of statistics applications that may help to streamline the analysis of data. Some of these tools are extremely sophisticated and powerful, mainly when coupled with current computers and intuitive graphical user interfaces (GUI) like Mac OS X http://www.apple.com/macosx. Yet, most of them come short of usability for the non-statistician. In other words, the users must know a lot of statistics to use and master them; they must know what to do with the data, and then use the application to accomplish their goals. But what about when the life scientists know little about statistics? Fortunately, since some years ago, there are some amazing applications that are built in such a way as to guide and even teach the users in the otherwise challenging statistics world. Two of such jewels are the computer applications StatMate http://www.graphpad.com/StatMate/statmate.htm and InStat 3 http://www.graphpad.com/instat/instat.htm from GraphPad. This comes to no surprise, when it is noticed that the founder of such company (Harvey J. Motulsky) http://graphpad.com/www/about.htm is also the author of the excellent book "Intuitive Biostatistics" (Motulsky, 1995) http://graphpad.com/www/book/book.htm. If you are a student or a researcher looking for an easy to read and understand statistics book, look no further. Contrary to other books, written by statisticians for statisticians, "Intuitive Biostatistics" teaches statistics to anyone, without requiring a previous strong mathematical background. The beauty of this book is that it represents a practical approach that focuses and emphasizes not on the mathematical calculations, but on the scientific problem, how to analyze it and how to interpret the statistical data generated. As an example of this fortunate pedagogic strategy, the book explains and deals with the "confidence intervals" (which are easier to understand by anyone), rather than the "P values" (which may be somewhat obscure by the non-statistician). The net result is that, eventually, it makes it easier to understand both statistical concepts by the reader. One of the questions that life scientists should ask themselves when designing experiments and before carrying them out is the number of data points required, as well as the power of the experiment to tell apart the different hypotheses. The amazing thing about StatMate is that it can answer such questions and teach the users statistics at the same time. The users can experiment with the different approaches, their advantages and disadvantages. And since all information is just built-in the application GUI, there is virtually no learning curve. Thus, using StatMate is a four-step process: 1. Choose the analysis. Depending on the goal, the users can select between sample size for a future experiment, or determine the power of a completed experiment. If the users need help, they can just click the button "Master the concepts of power and sample size" and the StatMate help window will pop-up. They can also click the "Learn" button at any time. Additionally, the users may select the experimental design between the five supplied, including comparison of means and hypothetical values (t test), survival curves (log-rank test) and proportions (chi-square). 2. Define experiment. Depending on the experimental design previously chosen, the users will enter the values for the requested statistical parameters, like standard deviation (SD), correlation coefficient (r), survival or success proportion, tabulation and significance level. If the users make a mistake, a warning window pops-up displaying -thus teaching- the legal ranges. Again, the button "Help me decide" will explain such concepts. 3. Choose power and N. Click any of the figures in the displayed "Difference between means that can be detected" table. 4. View report. A comprehensive report is shown, indicating the previous choices, the tradeoffs and the alternative options. There are also buttons to "Print", "Copy" and "Graph" to plot results (which requires the latest version of Prism http://www.graphpad.com/prism/Prism.htm from the same developer). Once the above is set, the experiments carried out and the data collected, the users may be interested on performing more "real" statistics with them. Here is where InStat excels. This extraordinary application works as a statistician coach, guiding the users in a six-step process of defining the data type, entering the actual data, choosing the best statistical test, and finally analyzing and displaying the results of the statistical analysis. Thus, InStat works in much the same way as the previously described for StatMate: 1. Kind of data. The users choose the kind of data to be analyzed. For that, they select a goal, like comparing means (or medians), regression and correlation, or contingency tables. Then they choose the data entry format between the ones supplied. Based on such selections, a description of the different available tests is shown on the right side of the InStat window. At any time, the users can click the buttons "Explain my choices" and "InStat Guide" (which also may pop-up with new selections or steps), to get comprehensive and useful explanations. 2. Data. The users enter the actual data in a grid table. They can click the "Arranging Data" and "Importing Data" buttons for explanations and help. 3. Statistics. When applicable, a list of different statistics values is given for the previously entered data. Clicking the "Interpret" and "Entering Mean&SD" buttons provide further assistance. 4. Choose test. The users answer the displayed questions. They can click the buttons "Help me choose" and "Selecting Columns" if needed, for further assistance and help. Based on the previous choices, the InStat wizard will choose a test and report about such selections. 5. Results. A comprehensive window of results is given, based on the previous selections. Clicking the "Checklist" and "What's Next?" buttons provide further explanations and help. 6. Graph. A plot is generated, using the analyses carried out before. Clicking the "GraphPad Prism" and "What's Next?" buttons display further information and assistance. When applicable, the buttons SD and SEM allow you to toggle the graph to show either the "mean and standard deviation", or the "mean and standard error", respectively. The button "What's Next?" explains how to print or export the results, view a graph, record notes or append the results to the notes window, analyze the same data with a different test, perform the same analysis on new data, start InStat again, and create an analysis template. The users may spend countless hours exploring their data, the InStat possibilities and, at the same time, learning statistics in an intuitive and pleasant environment that will gladly impress them. But do not take our words by granted. Read "Intuitive Statistics" if you can, go to the GraphPad web site, read the application information, browse the guided tours and -most importantly - download and test the fully functional demos of both StatMate and InStat. Put them to the test with your dada, and you will be gladly surprised. Chances are that whenever you want to design a new experiment or perform statistical tests on your data, you can not help but use such fantastic tools. Note, nevertheless, that although both StatMate and InStat are extremely intuitive and useful, they do not cover all aspects of statistics. If you need other tests or more powerful tools, then you have to look elsewhere, to either standard statistical packages (those hard to understand for non-statisticians), collaborate with or hire a statistician. But for the supported tests and analyses, both StatMate and InStat are excellent tools. Last but not least, thanks to the Mac code gurus (the programmers behind StatMate, InStat and Prism) at "Software MacKiev" http://www.mackiev.com for making this dream come true. Rating: 9.0 (4.5 out of 5 stars). Acknowledgements Supported by "Grupo PAI AGR-001" and "Grupo PAI TEP-169" of "Junta de Andalucía" (Spain). References Motulsky HJ (1995): "Intuitive Biostatistics". Oxford University Press (Oxford). Web: http://www.graphpad.com/www/book/book.htm. Motulsky HJ (2001): "Statistics for Biologists. The InStat Guide to Choosing and Interpreting Statistical Tests. GraphPad InStat Version 3 for Macintosh". GraphPad Software (San Diego). Web: http://graphpad.com. ### << Previous Next >> [ View All Perspectives ] |
|