I've tried running a formula similar to y ~ 1. This gives me just an intercept, but is the value that's shown equal to intercept estimate(without covariates) + 1, or just intercept estimate(without covariates)?
Any help would be appreciated.
$\begingroup$
$\endgroup$
5
asked May 5, 2016 at 16:48
An old man in the sea.
5,9825 gold badges33 silver badges75 bronze badges
1 Answer
$\begingroup$
$\endgroup$
1
When you do linear regression on only a constant, you will only get the intercept value, which is really just the mean of the outcome.
In R we have:
y <- rnorm(1000)
lm(y ~ 1)
# intercept = 0.00965
mean(y)
# Equal to 0.00965
The reason for doing it the regression way, rather than just computing the mean, is to get an easy standard error.
If have no idea what you mean with "intercept +1", but what you do get is the mean of the outcome, i.e. the expected value of y when you do not control for anything.
-
2$\begingroup$ Thanks Repmat. Also, another reason for doing this, is that some packages require the user to define a base model, e.g. 'BayesVarSel'. $\endgroup$An old man in the sea.– An old man in the sea.2016-05-05 17:16:50 +00:00Commented May 5, 2016 at 17:16
yand the constant1, how could it possibly even know about any covariates? $\endgroup$