Single and Quadratic function in one loop

[Paul Shlapa]

Dear Experts,

I need you advice on the possibility to loop 2 codes into one. One is quadratic function and another linear. The main problem is with local beta2`i'_`select' = _b[`select'2], which generates a macro for quadratic variable. It does nor allow the code to finish if used in a single function regression.
One has to somehow disable it once a single function regression runs.

The code below works fine, but i have to try to combine the loops.

//quadratic equation
foreach j in "Volume Volume2 `varlist_`i''" "TransferPrice TransferPrice2" {
    capture noisily reg GlobalPrice `j' [aw = total_SalesRevenue_EUR_per_Year] if groups == `i', nocons baselevels
        
    tokenize "`j'"
    local select "`1'"    
        
    local beta1`i'_`select' = _b[`select']
    local beta2`i'_`select' = _b[`select'2]
        
    di "`PumpName'"
    di "quadratic equation is `beta1`i'_`select''*x + `beta2`i'_`select''*x^2"
    local g`i'_`select_quadratic' = "function y = `beta1`i'_`select''*x + `beta2`i'_`select''*x^2, range(`m_min`i'' `m_max`i'')"
    di "`g`i'_`select_quadratic''"
}

//linear equation
foreach j in "Volume `varlist_`i''" "TransferPrice" {
    capture noisily reg GlobalPrice `j' [aw = total_SalesRevenue_EUR_per_Year] if groups == `i', nocons baselevels
        
    tokenize "`j'"
    local select "`1'"    
        
    local beta1`i'_`select' = _b[`select']
            
    di "`PumpName'"
    di "linear equation is `beta1`i'_`select''*x"
    local g`i'_`select_linear' = "function y = `beta1`i'_`select''*x, range(`m_min`i'' `m_max`i'')"
    di "`g`i'_`select_linear''"
}
}

Thank you

[Jared Greathouse]

Give a data example please. As well as, be more specific about what you want, please. The most I understand is the you want to square a variable.... I think.

[Daniel Schaefer]

What about something like this:

// quadratic and linear combined
foreach j in "Volume `varlist_`i''" "TransferPrice" {
    
    tokenize "`j'"
    local select "`1'"   
    
    capture noisily reg GlobalPrice `j' `select'2 [aw = total_SalesRevenue_EUR_per_Year] if groups == `i', nocons baselevels
    
    local beta1`i'_`select' = _b[`select']
    local beta2`i'_`select' = _b[`select'2]
    
    di "`PumpName'"
    di "quadratic equation is `beta1`i'_`select''*x + `beta2`i'_`select''*x^2"
    local g`i'_`select_quadratic' = "function y = `beta1`i'_`select''*x + `beta2`i'_`select''*x^2, range(`m_min`i'' `m_max`i'')"
    di "`g`i'_`select_quadratic''"
    
    capture noisily reg GlobalPrice `j' [aw = total_SalesRevenue_EUR_per_Year] if groups == `i', nocons baselevels
    
    local beta1`i'_`select' = _b[`select']
 
    di "`PumpName'"
    di "linear equation is `beta1`i'_`select''*x"
    local g`i'_`select_linear' = "function y = `beta1`i'_`select''*x, range(`m_min`i'' `m_max`i'')"
    di "`g`i'_`select_linear''"
}

Above, I use your tokenize, then `select'2 syntax to add what I assume is the quadratic term to the model dynamically.

[Paul Shlapa]

Done, I have used a slightly different approach that the one that Daniel proposed (did not solve the problem of local beta2`i'_`select' = _b[`select'2]
Now it only defines local, if select2 is present.

//quadratic equation
foreach j in "Volume" "Volume Volume2" "TransferPrice" "TransferPrice TransferPrice2" {
    
    tokenize "`j'"
    local select1 "`1'"
    local select2 "`2'"
    
    capture unab ``j'' : ``var''
    local variable_count_equation = wordcount("`j'")
    
    capture noisily reg GlobalPrice `j' [aw = total_SalesRevenue_EUR_per_Year] if groups == `i', robust baselevels

forvalues m = 1/`variable_count_equation' {
    
    local beta_`i'_`select`m'' = _b[`select`m'']
}    

    di "`PumpName'"
    di "quadratic equation is `cons`i'_`select'' + `beta1`i'_`select''*x + `beta2`i'_`select''*x^2"
    local g`i'_`select_quadratic' = "function y = `cons`i'_`select'' + `beta1`i'_`select''*x + `beta2`i'_`select''*x^2, range(`m_min`i'' `m_max`i'')"
    di "`g`i'_`select_quadratic''"
}

Thanks everyone for your time