I know that you can hold your data as a "matrix" but I am not so sure that you have something that is useful to think of as a "table for categorical analyses."
On 29 Oct 2003 05:09:49 -0800, [EMAIL PROTECTED] (Cave) wrote: > I've got a 4*2*8 table where: > - The first dimension is the year of the data (4 values: 1999-2002). This is okay. Dimension one is the year; at a first approximation, I assume that you could *collapse across year* in order to simplify your presentation. You probably have to assume that the data are independent for the different years -- unless you do have raw data for every individual for all years (something not mentioned, but very useful if you have it). > - The second dimension holds number of occupational injuries and total > number of workers (a worker can experience more than one injury in the > same year, but each injured worker is counted appart). For simple counts, the second dimension should be, Persons who were: Injured versus not-injured (in this year). If this is a table of *counts*, then *multiple-injuries* probably belongs somewhere else. You could (it seems) figure out the average injuries-per-person. Is there any new and useful information in the data about multiple injuries? or is that just a complication in trying to describe the data. > - The third dimension represents groups according to experience at > work (less than 6 months, from 6 to 12, ... etc., a total of 8 > groups). So, the counts are broken down by 'experience' -- Here is a definitional problem. Is that, "experience at the time of the accident" or at the beginning of the year (or its end)? Where are people counted who started work, and quit, but never had any accident? - There will be *some* problem in identifying, for instance, how many people were exposed to the job for their "first month". Just off hand, it sounds like an interesting problem of data-collection design, and subsequent analysis. If the design was not subtle enough, you might end up guessing, at best. > I'm interested in analizing the relationship between the second and > third dimensions, that is, the relationship between injuries by worker > and experience. > My question is �wich is the appropiate analysis for my study? > I would thank very much any help, and also if you have references of > this kind of analysis. "occupational epidemiology". Cohort studies. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html "Taxes are the price we pay for civilization." . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
