I thank Dennis, Gordon, and Jay for their kind response. Apparently, my question is a lot more confusing than I thought, to the effect that Bob Wheeler felt that I intentionally over formulated the problem. I didn't, Bob. Here I'd like to give a precise statement of the problem at hand, and I would like to invite further suggestions. Given the suggestions of Dennis, Gordon, and Jay, I have a feeling that maybe what I am trying to do asks too much from the data. You'll be my judge.
I have a survey dataset which has the information of the respondents' actual incomes in year 1, and the survey question also asks the respondents to rate their expected income in the following year after the interview. Their answers are recorded in the scales of "increase by less than 10%", "increase by 10% to 20%", etc.. What I am trying to obtain is a measure of "income uncertainty" or "income variability" perceived by the respondent at year 1, which I labeled "expected income variance". A bad term, may be. So, I stated the problem in my previoius posting as if the individual has income 100 in year 1, and expected to have income between 110 and 120 in year 2. Then I ask how to calculate the "income variance" (confusing term). Ideally, the income variance should measure the variability of incomes in year 1 and year 2. Now, if I have the individual's "actual income" in year 1, 2, 3, ..., I would have no problem calculating the income variance: simply calcualting the variance of those numbers. What I could not figure out is that in the case I have only 2 observations and one of which is an "expected" "range" of income, how could I calculate the corresponding income variance? Many thanks in advance, Eddy . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
