For comments / advice: a complex ado file for a neophyte

[Basharat Hussain]

Kindly check this code? I am trying to calculate Median Absolute Deviations. What could be done better? Any errors in logic / calculations? Is the use of global macros justifiable here? The code has been adopted from Dr. Nicholas J. Cox's moments.ado, with his kind permission.

program mads, byable(recall)   
        version 9.1
        syntax [varlist] [if] [in] [aweight fweight] /// 
        [, Matname(str) Format(str) ALLobs variablenames by(varlist) * ]

    qui { 
        ds `varlist', has(type numeric) 
        local varlist "`r(varlist)'" 

        if "`allobs'" != "" marksample touse, novarlist 
        else marksample touse 

        count if `touse' 
        if r(N) == 0 error 2000 

            local nct : word count `varlist'
    
        if "`by'" != "" { 
            if `nct' > 1 { 
                di as err ///
                "by() cannot be combined with `nct' variables"
                exit 198 
            }
            tempvar group 
            egen `group' = group(`by') if `touse', label 
            su `group', meanonly  
            local nct = r(max) 
        }     
        else tokenize `varlist' 

            if `nct' > _N {
                    preserve 
                    set obs `nct'
            }

            tempvar a id n mean SD median mad which
            tempname mylbl 

                gen long `which' = _n
                compress `which'
                
        gen `n' = "" 
        label var `n' "n" 
        
        foreach s in mean SD median mad { 
            gen double ``s'' = . 
                    label var ``s'' "`s'"
        }     
            
            if "`matname'" != ""  mat `matname' = J(`nct',5,0) 
                
            forval i = 1/`nct' {
            if "`by'" != "" { 
                su `varlist' [`weight' `exp'] ///
                if `touse' & `group' == `i', detail     
                global gillid = `varlist'
                global gillN = r(N)
            }     
            else   { 
                        su ``i'' if `touse' [`weight' `exp'], detail
                        global gillid = ``i''
                        global gillN = r(N)
             }
                
                        replace `n' = string(r(N)) in `i'
                        replace `mean' = r(mean) in `i'
                        replace `SD' = r(sd) in `i'
                        replace `median' = r(p50) in `i'
                        /* here gores replace `mad'*/
            
                        tempvar devi 
                        ge `devi' = abs($gillid - `median') in `i'
                        su `devi' in `i', de        
                        replace `mad' =  r(p50) * 1.4826 in `i'
                        
                        
                    
                    if "`matname'" != "" {

                mat `matname'[`i',1] = $gillN
                mat `matname'[`i',2] = r(mean)
                mat `matname'[`i',3] = r(sd) 
                mat `matname'[`i',4] = r(p50)
                mat `matname'[`i',5] = `mad' 
                
                    }
            if "`by'" != "" { 
                local V = trim(`"`: label (`group') `i''"')
                local rownames `"`rownames' `"`V'"'"' 
                display _newline(2) "rownames = "`rownames'""
            }
            else {
                local V = trim(`"`: variable label ``i'''"')  
                if "`variablenames'" != "" | `"`V'"' == "" { 
                    local V "``i''" 
                }     
            }    
            label def `mylbl' `i' `"`V'"', modify 
        }

            if "`matname'" != "" {
                    mat colnames `matname' = n mean SD median mad
            if "`by'" != "" { 
                capture mat rownames `matname' = `rownames' 
                if _rc { 
                    numlist "1/`nct'" 
                    mat rownames `matname' = `r(numlist)' 
                }
            }    
                    else mat rownames `matname' = `varlist' 
            }

            label val `which' `mylbl'
        if "`by'" != "" label var `which' "Group" 
        else if "`allobs'" != "" label var `which' "Variable" 
            else label var `which' "n = $gillN"

            local fmt "format(%9.3f)"
        
        if "`format'" != "" { 
            tokenize `format' 
        
            if "`4'" != "" { 
                tempvar smad 
                gen `smad' = string(`mad', "`4'") 
                label var `smad' "mad" 
                local mad "`smad'" 
            }     
        
            if "`3'" != "" { 
                tempvar smean 
                gen `smean' = string(`mean', "`3'") 
                label var `smean' "mean" 
                local mean "`mean'" 
            }     
        
            if "`2'" != "" { 
                tempvar sSD 
                gen `sSD' = string(`SD', "`2'") 
                label var `sSD' "SD" 
                local SD "`sSD'" 
            } 
        
            tempvar smean 
            gen `smean' = string(`mean', "`1'") 
            label var `smean' "mean" 
            local mean "`smean'" 
        }     
        
        
        if "`allobs'`by'" != "" local shown "`n'" 
    }    
            
    tabdisp `which' if `which' <= `nct', ///
        c(`shown' `mean' `SD' `median' `mad') `options' `fmt'
       
end

[Richard Williams]

Do you have an ability to reproduce known results, e.g. from a published article; or maybe run something in other software (SAS, R, SPSS) where this is already programmed. The first concern is always accuracy. You can fine tune for ease of use, error checking, and nice displays once you know the results are right.