I can't help it. the last paragraph in this post absolutely _demands_ a response.
Wuzzy wrote: > > You made a model with the "exact same exposure in different units", > > which is something that no one would do, > > Hehe, translation is don't post messages until you've thought them > through. > > Anyway, turns out that the answer to my question is "No".. > Multicollinearity cannot force a correlation. It turns out that ONE > of the variables *was* correlated With R^2=0.45 and so > multicollinearity had no effect on overall R^2. > > I'm sure no-one is interested in my data as it has nothing to do with > statistics, my subject of interest is not statistics.. but i need to > learn it as a tool.. Dear Wuzzy, In two short sentences, you have expressed the fundamental issues of those who claim their "subject is not statistics." So long as you try to separate 'statistics' from your specific technology, you will not develop much of either. House builders do not spend much time concerned with their hammers or their nails. Yet they are sufficiently concerned that (in the USA) they buy expensive nail guns and special nail packages so they can build those houses faster and better. They use roofing nails to hold the roof shingles in place, finishing nails to hold the interior trim in place, and they carefully know the differences between them. In resolving a technical product performance problem, which is what I largely do with my statistics, I have to carefully decide what I am going to measure, how it will be measured, and how I will analyze it (crunch the numbers). Many people believe that this last step equals statistics. They neglect that the analysis methods depend on those first two items. They often neglect that each of those numbers I crunch _means_ something. They have units. They relate back to what was measured. The statistical analysis in my view is mostly concerned with detecting and quantifying the relationships between the different things and conditions which were measured. Thus, without the statistics, you have no technology; without the technology you have no statistics. You cannot relegate one of them down to the level of 'tool.' Down that path lies the perennial question, 'which equation should I use,' which begs its own questions. In your specific case, it appears that you tried to do a multiple regression using one response (dependent variable) and three factors (independent variables). But the three factors were actually transformations of the same variable. Since you said they were in different units, the transformations were probably linear. If you tried to do a full multiple regression on this data in this manner (3 factors), I'm surprised that the software did not warn you it had found a singular matrix, or at least that it had tried to divide by zero. Perhaps you made the conversions on a hand calculator, so small rounding errors kept the matrix that is inside the analysis from blowing up (inward!?:) on you. In any case, discovering that a linear transformation of data produces radically different r^2 values should be a warning that something is amiss, and it is time to think more carefully about exactly what digits are being pushed around the screen. Those numbers _mean_ something, remember :) And so does the math of the equations we select. A correlation and/or linear or polynomial regression analysis with one response and one factor would probably be more technically valuable, for your data, as best I can see from here. As for interest in your data, I can say that I would like to see it, as an example I can use for students. I need to collect real data from many different technologies - industrial, business, medical, social sciences, etc. - in order to relate the topic of 'statistics' to the areas of interest to them. I will be happy to share the write up with you, especially if you are willing to correct any errors in the technology which I am likely to make. I am also careful not to slam even the fictitious people who appear in them. After all, it takes an expert to make a good hammer, and an expert to make a good house. They need each other, just as 'statisticians' need 'technology experts.' Cheers, Jay -- Jay Warner Principal Scientist Warner Consulting, Inc. 4444 North Green Bay Road Racine, WI 53404-1216 USA Ph: (262) 634-9100 FAX: (262) 681-1133 email: [EMAIL PROTECTED] web: http://www.a2q.com The A2Q Method (tm) -- What do you want to improve today? ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at http://jse.stat.ncsu.edu/ =================================================================
