How to interpret a t-test?

[Giovanni Abronzino]

Hi, I'm new on this forum and, essentially, i'm not that good in statistics but i need to run a project work on multiple regression model because of an exam. I have got a dataset where the dependent variable is a discrete one "quality of wine", and 10 indipendent variables. I ran t-tests for all the predictors and i don't know how to intrepet them. The central problem is the hypotesized mean, i really don't know what is the value i should choose. So i really need help! I attached results coming from three t-tests on three of the dependent variables.
PS. I really need some explanation, but please, i haven't studied advanced statistics, so please speak as clear as possible!! Thank you

[Nick Cox]

If you feed a t test a value very close to the observed mean, then it's fairly unsurprising that the one-tailed P-values are each about 0.5.

More crucially, there is no obvious call for a one-sample test here unless you have a substantive hypothesis that the mean could be or should be a certain value.

We have a policy of not supporting homework or work for assignments, and this seems close to the boundary.

[Carlo Lazzaro]

Giovanni.
as your question related to a graded assignment, please see https://www.statalist.org/forums/help#adviceextras, #4.
That said, why using -ttest- when you were assigned a multiple regression?

PS: crossed in the cyberspace with Nick's reply, with whom I fully agree.

[Giovanni Abronzino]

Well, i apologize for not reading the policies. This is a crazy situation in which i didn't have a course neither any book where i could study, because the professor simply didn't want to give online lessons. I just have some very bad guidelines which i had to interpret, it is difficult for me because i don't have enough background studies! In these guidelines i was told to ran t-tests to check the statistical significance of variables, nothing more.

[Marcos Almeida]

Well, this is very sad. With regard to guidelines, a great starting point is the Stata Manual.

That said, to run t-tests in such scenario, you must have the predicors as binary variables, but it won't provide a great analysis, as Carlo and Nick pointed out.

But something like - ttest yvar, by(predictor) - can do the trick.

The best starting point is surely typing - help ttest - in Stata's command window.