I have found the R program of the linear regression for the consumer price index of a country, expressed in terms of years and quarters division. The formula obtained is:
so that mathematically it would be something like:
which is a polynomial linear model, for what a think. The correlation is positive and in the graph the slope is positive so it is a growing line. One problem that I have here is the measure of the intercept, according to the authors and the R program, it is a value of -7619.39, why is that? it is pretty big for the cpi values that are from 160 to 174.
Now there is another example that wants to predict the bodyfat based on the measures of the waist, hip, etc.
The formula will be:
myFormula <- DEXfat ~ age + waistcirc + hipcirc + elbowbreadth + kneebreadth
and the authors use a generalized linear model like:
bodyfat.glm <- glm(myFormula, family = gaussian("log"), data = bodyfat)
the question that I have is why the produced model it is not a polynomial case like the one of the cpi? for me both seem the same; and here is using a family of gaussian distributions and R for the simple linear regression uses a normal distribution which is the same. What am I missing here? Thanks