Independent t-test for differences other than 0

[Amanda Ode]

Dear Stata Experts,

I would like to run the independent t-test for the following dataset and want to test the null hypothesis for a difference of 300 instead of the default 0.
The command I use so far is ttest "Outcome, by(Type) unequal"

* Example generated by -dataex-. To install: ssc install dataex
clear
input int Type double Outcome
101 2997.5198380566803
101 2937.7315789473687
101 2955.9720647773283
101  2956.985425101215
101 2974.2125506072875
101 2961.0388663967615
101 2971.1724696356277
101  2976.239271255061
101 2966.1056680161946
101  2948.878542510122
101  2930.638056680162
 70  2598.276923076923
 70             2566.8
 70  2603.046153846154
 70  2646.923076923077
 70 2438.9846153846156
 70 2588.7384615384617
 70 2579.2000000000003
 70 2624.9846153846156
 70  2595.415384615385
 70 2586.8307692307694
 70 2625.9384615384615
end

Is there a way of doing that?

Thank you in advance.
Best,
Amanda

[Nick Cox]

Where does the 300 come from? If it's suggested by the data, then it seems to me that the real issue is a confidence interval for the mean difference.

[Amanda Ode]

The command I use so far is
ttest Outcome, by(Type) unequal

[Amanda Ode]

Dear Nick,
We have a specification limit for the maximum allowed difference between the means and that is, for this case, 300. I just want to test the following hypothesis:

H₀: The difference (P=0.95) between results obtained at pressure 1 and pressure 2 is significantly less than 300.
H₁: The difference between measuring at pressure 1 and pressure 2 is significant.

Hope that helps!

[Amanda Ode]

Pressure=Type in the code above.

[Richard Williams]

As a wording matter, I don't think the word "significant" should be in the phrasing of your hypotheses. It should be more like

H0: The difference between results obtained at pressure 1 and pressure 2 is <= 300.
H1: The difference between results obtained at pressure 1 and pressure 2 is > 300

As for testing -- as Nick says, just look at the confidence interval, which in this cases ranges from -412.2304 to -337.107. -300 is not in the CI so reject the null.

If this is meant to be a one-tailed test, which is what it sounds like to me given the way you phrase things, you can also do something like

clonevar XOutcome = Outcome
replace XOutcome = XOutcome + 300 if Type == 70
ttest XOutcome, by(Type) unequal

[Amanda Ode]

Thank you Nick and Richard!
I was curious about the possibility of having the number 300 in the command. But as you say, I'll continue interpreting the results by using the CI. I will try your suggestion as well Richard, just for learning more about Stata
Best,
Amanda