最近在计算产业集聚的衡量指标，在理解EG指数上遇到了2个问题，希望各位好心的前辈可以予以解答：
1. 最原始的EG指数是把一系列“行业-地区”的数据加总到行业层面，就像附件里图片中的公式所展示的，所以我理解的EG指数应该是行业层面的一个指标，仅有行业上的variation。但最近读了一些文章，好多在跑回归的时候都是用的"地区-行业"层面的EG指数，这个地区层面的variation是哪里来的呢？是在计算出的行业层面的EG指数进行平均了还是直接就计算出来了？关于这一点我一直没有理解。

2. 关于egi命令，egi命令跑出来的EG指数是“行业-地区”层面的，很好奇为什么会有一个地区层面的variation而不是仅仅只有行业层面的variation？而且看了egi编程的具体说明，里面的计算过程也确实是把地区层面的各项指标进行了加总，加总到了行业层面，以下是egi命令的原文解释：

*! version 9.2  20oct2015 Author : Jay Dev Dubey *! input recieved by Pritam Datta,  (PHFI)*The Ellison and Glaser's index; cap program drop egiprogram define egiversion 9syntax varlist(max=1), in(varlist) st(varlist) index(name)  [,mult(varlist) by(varlist)]qui sum `varlist'        if r(min)<0{                display as error "Negative values in varlist"                exit 198                    }if "`index'"!="EG" & "`index'"!="HHI" & "`index'"!="GI" {di as err "Please select appropriate concentration index"        exit 198        }if "`mult'"=="" {tempvar multqui gen `mult' = 1}if "`by'"=="" {tempvar byqui gen `by' = 1 }tempvar sipc sipc1 Sist spc spc1 Sst Git pc pc1 pc2 Hit*****Group wise total worker********tempvar  alpha1 alpha2 alpha3 alpha4qui egen `alpha1' = sum(`varlist'*`mult') , by(`in' `st' `by')qui egen `alpha2' = sum(`varlist'*`mult') , by(`in' `by')qui egen `alpha3' = sum(`varlist'*`mult') , by(`st' `by')qui egen `alpha4' = sum(`varlist'*`mult') , by(`by')*****Direct Ratios for Ss and Gi********tempvar alpha12 alpha34qui bysort `by': gen `alpha12' = `alpha1'/`alpha2'qui bysort `by': gen `alpha34' = `alpha3'/`alpha4' ********State Wise and total Ss********tempvar alpha5 order1 Ssqui bysort `by' :  gen `alpha5' = (`alpha34')^2 qui  bysort `st' `by' : gen  `order1' = _n qui replace `order1' = 0 if `order1' > 1qui replace `alpha5' = `alpha5'*`order1'qui egen `Ss' = sum(`alpha5') , by(`by')*******Calculation of Gi**********tempvar Gdev order2 Gi GIqui bysort `by' : gen `Gdev' = (`alpha12' - `alpha34')^2 bysort `in' `st' `by' : gen `order2' = _n qui replace `order2' = 0 if `order2' > 1qui replace `Gdev' = `Gdev'*`order2' qui egen `Gi' = sum(`Gdev') , by(`in' `by')qui gen `GI' = `Gi'/(1-`Ss')*su `GI'********************************************Herfindhal Style Index********tempvar her1 her2 her3 HIqui gen `her1' = `varlist'/`alpha2'qui gen `her2' = (`her1')^2qui gen `her3' = `her2'*`mult'qui egen `HI' = sum(`her3') , by(`in' `by')*******************************Final Step For GE index calculation***********if "`index'"=="EG" {qui gen `varlist'_GEI = (`GI'-`HI')/(1-`HI')label var `varlist'_GEI "Ellison and Glaser's Index" }if "`index'"=="HHI" {qui gen `varlist'_HHI = `HI' label var `varlist'_HHI "Herfindhal style measure of plant level concentration" }if "`index'"=="GI" {qui gen `varlist'_GI = `Gi' label var `varlist'_GI "Sum of square deviation of industry - state employment share" }end


我本人感觉有点不太明白，明明这里计算过程中生成的的GI、HI都是行业层面的，不知道为什么计算的结果会多出地区层面的variation，哪位前辈可以帮忙解答一下吗？感激不尽！