Possible Duplicate:
When is it ok to remove the intercept in lm()?
I have some regression output from R:
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -28.84775 29.71148 -0.971 0.341
sqft 0.17091 0.01545 11.064 2.48e-11 ***
lotsize 6.77770 1.42129 4.769 6.19e-05 ***
baths 15.53470 9.20827 1.687 0.104
---
The intercept (and baths) looks useless, so I ran confint() to confirm it:
2.5 % 97.5 %
(Intercept) -89.9205710 32.2250754
sqft 0.1391546 0.2026609
lotsize 3.8561912 9.6992184
baths -3.3931660 34.4625741
And what do you know, it is. Now I am stuck on what to do here. I see that the intercept can have a zero slope, so can I treat the intercept like a predictor and drop it?
After taking baths out and running the regression again, the intercept is still coming up as statistically insignificant:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -23.20585 30.51527 -0.760 0.453566
and
2.5 % 97.5 %
(Intercept) -85.8180248 39.4063179
Any ideas?
