T-test over regression constants instead of whole data set

[Daniela Bonilla]

Hi,

I did a loop regression and now want to compute a t-test only on the constants being calculated. My current code is calculating the t-test over the whole data set. Any help?

use "Data - RLBOS Stock and FF Master1.dta", clear

merge m:m DataDate using "Benchmark Daily.dta"

drop if _merge == 1 | _merge == 2
drop _merge

sort GVKEY DataDate DateCompletedIssue

bys GVKEY seconddate: egen benchmarkyrret = mean(ValWeightRet)

sort GVKEY DataDate 
by GVKEY (DataDate): gen dealexcessreturns = monthlyreturn - benchmarkyrret 

bys GVKEY seconddate: egen marketexcessreturns = mean(mktrf)
     
sort GVKEY 
egen group_id = group(GVKEY)

** Run the CAPM regression for the first group range
encode GVKEY, gen(GVKEY_num)
estimates clear
levelsof GVKEY_num, local(groups)

local start_group1 1
local end_group1 88

scalar sum_alpha1 = 0  // Initialize a sum variable for alpha1
scalar count_alpha1 = 0  // Initialize a count variable for alpha1

local constants = 0

bysort GVKEY seconddate: keep if _n==1 
bysort GVKEY: drop if _n==1 

gen alpha_values = .

foreach group of local groups {
    if `group' >= `start_group1' & `group' <= `end_group1' {
        regress dealexcessreturns marketexcessreturns if GVKEY_num == `group', robust
        estimates store capm_`group'
        scalar alpha1 = _b[_cons] - (rf + benchmarkyrret * marketexcessreturns)
        display "`group' alpha1: " alpha1
        
        replace alpha_values = alpha1 if GVKEY_num == `group'
        
        // Update the sum and count for alpha1
        scalar sum_alpha1 = sum_alpha1 + alpha1
        scalar count_alpha1 = count_alpha1 + 1
    }    
    
}

scalar constants = sum_alpha1 / count_alpha1

ttest alpha_values = 0 if !missing(alpha_values)
signrank alpha_value = 0