Test of intercept - Panel regression

[Leandro LF]

Dear experts,

I have been trying to test if the intercept (alpha) of a panel regression is significant higher than zero. I have collected data of 700 investment funds for the last 3 years, and would like to know if, overall, the constant (intercept) is positive.
I tried testparm, but I believe that this function just tells me if the intercept is equal or different from zero.
Does anyone know which test can I run to know if the intercept is positive?

Thank you all!

[Attaullah Shah]

If you collect the intercepts and place them in one column, then you can apply ttest on it. ttest varname==0 produces all the three test statistics for mean of a variable: i.e. it is equal to zero, it is greater than zero, and it is less than zero. See the following example.

clear
set obs 100
gen intercept=rnormal()
ttest intercept==0


. ttest intercept==0

One-sample t test
------------------------------------------------------------------------------
Variable |     Obs        Mean    Std. Err.   Std. Dev.   [95% Conf. Interval]
---------+--------------------------------------------------------------------
interc~t |     100   -.0132365    .0947744    .9477437   -.2012894    .1748164
------------------------------------------------------------------------------
    mean = mean(intercept)                                        t =  -0.1397
Ho: mean = 0                                     degrees of freedom =       99

    Ha: mean < 0                 Ha: mean != 0                 Ha: mean > 0
 Pr(T < t) = 0.4446         Pr(|T| > |t|) = 0.8892          Pr(T > t) = 0.5554

Since the data is randomly generated, the test statistics show 44.46% confidence level that the value of intercept is less than zero.

[Leandro LF]

Mr. Shah, thank you very much for the answer.
Nevertheless, I have collected the daily returns of 700 funds for the last 3 year, which means that I have more than 500.000 intercepts.
How do I collect than all and place in one column?

I see you are a professor of finance. If it helps, what I would like to test is the Jensen's alpha for the hedge fund industry in Brazil.

Thank you again.

[Attaullah Shah]

For collecting alphas in one column, you can use statsby, see

help statsby

Concerning the second question, we often do not stop at just testing the significance of alphas. In fact, testing for significance is the first step in a more formal analysis. We usually try to investigate whether funds' abnormal returns (alphas) are related to fund characteristics (see series of papers written by Elton and Grubber on this topic)

[Sebastian Kripfganz]

Please see the following Stata FAQ on one-sided tests:
The results from estimation commands display only two-sided tests for the coefficients. How can I perform a one-sided test?