one-side test

[Francis Czarnowski]

Hi All,

I have a quick question.

I aim at testing a null hypothesis of kind:
H0: b1+b2+b3+b4<0
H1: b1+b2+b3+b4>=0

In general, I wonder how to perform a one-sided, rather then two-sided test.

In case of the later I use test "b1+b2+b3+b4=0", but I have no idea how to specify the other test.

Thank you for your help.

Best,
Francis

[Carlo Lazzaro]

Francis:
see https://www.stata.com/support/faqs/s...-coefficients/

[Fei Wang]

Francis, if you do the test after regress, then you may use lincom to calculate the estimate of b1+b2+b3+b4. For example, if b1-b4 are the coefficients for x1-x4, then

reg y x1 x2 x3 x4
lincom x1 + x2 + x3 + x4

In the output, you will see a two-sided p value, and dividing it by 2 will give you the p value for one-sided test.