I work for a state government's agency that collects water quality data and often get requests for data by graduate students. (My organization collects A LOT of data.) Even though we provide the data to the students we often get many follow up questions on how to analyze the data, etc.
I am a firm believer in graphing data before doing any analyses. Error checking is essential, and in my opinion often overlooked. Graphing data is now-a-days a readily available way of identifying outliers, changes in analytical methodologies (e.g reporting limits). Our data (as most data) should be screened for inappropriate values (e.g. pH of 86 instead of 8.6, as pH can only range from 0 to 14). These error checking procedures do not take long. I use a statistical package called JMP (from SAS) and plot the concentrations of MANY parameters over time (this takes minutes). The resulting graphs (although crude) are often very informative and I can visually see problems, which are often corrected easily. Clearly if anyone is doing any data summaries, simple statistics or complex statistics on any one else's data, then error checking and graphing are (in my opinion) essential (they are even important using your own data). Looking at the data, getting a 'feel' for the data are really important. If any of you teach statistics, don't minimize error checking (but you don't have to spend hours on the topic either). ----- Original Message ----- From: "Esa M. Rantanen" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Cc: "Anna Cianciolo" <[EMAIL PROTECTED]> Sent: Monday, November 03, 2003 4:22 PM Subject: [edstat] Teaching statistics > Dear All, > > I have been exasperated by a particular deficiency (in my view) of > many graduate students in dealing with their data, namely, delving > into often complex statistical analyses without first having a good > 'feel' of where the data came from and what they should be looking > for. I was recently 'venting' my frustration to a colleague, who > will be teaching a two-course graduate stats 'package' in the near > future. We discussed ways to impress the importance of 'looking at > the data' on the students. I would like to pose the same questions > to members of this list; specifically, > > (1) how do you rate the importance of exploratory and (in particular) > graphical analysis of data prior to doing inferential statistics, and > > (2) how do you (or, would) incorporate these aspects of statistics > into your teaching? > > I am looking forward to your insights into these questions. > > Best, > > Esa > -- > ------------------------------------ > Esa M. Rantanen, Ph.D. > Assistant Professor > University of Illinois at Urbana-Champaign > Institute of Aviation, Aviation Human Factors Division > Willard Airport-One Airport Road, Q5, MC-394 > Savoy, IL 61874 > Tel. 217-244-8657 (AHFD) > Tel. 217-244-7397 (Psych.) > Tel. 217-373-8276 (Home) > Fax 217-244-8647 > e-mail: [EMAIL PROTECTED] > url: http://www.aviation.uiuc.edu/new/html/ARL/Esa_Rantanen.html > ------------------------------------ > . > . > ================================================================= > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > ================================================================= > . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
