No test can tell you if a variable is continuous or discrete. If you try to fit a continuous distribution to a discrete variable, you will either get an error or get a continuous distribution. Suppose you have data which are clearly discrete: Here are some count data I recently used (the vector is for number of people responding 0, 1, 2, 3, 4, 5, and 6 or more)
in R I wrote x <- c(1630, 3340, 663, 269, 114, 46, 226) if I now try to plot a density for this with plot(density(x)) I get what looks like a smooth curve. A variable is discrete if it can only take on a limited number of values; it is continous if it can take on any of a continuum of values. In a certain sense, all varaibles, as reported, are discrete: Measurements are limited to the accuracy of the reports; or, in any case, by the quantal nature of reality. Take weight. Ask someone how much they weigh. Whether or not they report accurately, they are apt to give a number of pounds (not pounds and ounces, not grams). Further, they are much more likely to say a round number (180, not 182) The questions are, "How can I best analyse these data?" and "Are they close enough to continuous?" I have not seen any research on these issues, but an intuitive feel is that a variable which can take on any integer between 0 and 100000 is so close to continuous as makes no difference. HTH Peter Casey wrote: > Hi, i needed some help in determining if some data i have are discrete > or continuous, i know this sound silly, but if someone can provide the > answer with some relevant proof, it will be much appreciated!!! I have > a string of data that thier values ranges from 0 to 100000 or more, > and values do not occur more than once, it looked discrete in the > first place but when i use a statistic software to fit a distribution, > the result is continuous, same happened when i do it manually using > conventional distribution fitting and testing method methods > (histogram, summary statistics, chi-square test and Kolmogorov-Smirov > test). So are these data discrete? or continuous? Is there other ways > or methods i can test the data? Will anyone please help, thank you for > your time!! Peter L. Flom, PhD Assistant Director, Statistics and Data Analysis Core Center for Drug Use and HIV Research National Development and Research Institutes 71 W. 23rd St www.peterflom.com New York, NY 10010 (212) 845-4485 (voice) (917) 438-0894 (fax) . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
