Running Sum Matrix

[Cindy Linwood]

Hello, I am new to Mata and how do you calculate a running sum matrix?

[Komal Jain]

Hi,

I am trying to run a loop where an individual has qty produced against variables 'qty_produce_kg' , 'yield' , 'dist_yr', 'subgrp'.

The subgrp here is a list of indicators such as age1 age2 etc.

I have tried to run the following loop, but it shows blank matrix:

set trace on

set matsize 800

matrix prel = J(33,25,.)
matrix rownames prel = age1 age2 age3 male female open obc scst openlease lease_land_yn fpc_subsidy_project subsidy_project_75pc subsidy_project_25pc subsidy_project_50pc nosubsidy intercropping monocropping preplanting_full preplanting_partial preplanting_none planting_full planting_partial planting_none afterplanting_full afterplanting_partial afterplanting_none
matrix colnames prel = Samastipur_2017 Samastipur_2018 Samastipur_2019 Samastipur_2020 Saran_2017 Saran_2018 Saran_2019 Saran_2020 Siwan_2017 Siwan_2018 Siwan_2019 Siwan_2020 EC_2017 EC_2018 EC_2019 EC_2020 WC_2017 WC_2018 WC_2019 WC_2020 tot_2017 tot_2018 tot_2019 tot_2020


local row A

foreach crop of local cropname {

local cropname blackgram chickpea fawabean greengram kidneybean lentil pea pigeonpea
local subgrp age1 age2 age3 male female open obc scst openlease lease_land_yn fpc_subsidy_project subsidy_project_75pc subsidy_project_25pc subsidy_project_50pc nosubsidy intercropping monocropping preplanting_full preplanting_partial preplanting_none planting_full planting_partial planting_none afterplanting_full afterplanting_partial afterplanting_none
local distyr Samastipur_2017 Samastipur_2018 Samastipur_2019 Samastipur_2020 Saran_2017 Saran_2018 Saran_2019 Saran_2020 Siwan_2017 Siwan_2018 Siwan_2019 Siwan_2020 EC_2017 EC_2018 EC_2019 EC_2020 WC_2017 WC_2018 WC_2019 WC_2020 tot_2017 tot_2018 tot_2019 tot_2020


foreach d of local distyr {

quietly summ qty_produce_kg_`crop' yield_`crop' if dist_yr == `d'

foreach s of local subgrp {

quietly summ qty_produce_kg_`crop' yield_`crop' if dist_yr == `d' & subgrp == `s'

egen concat = concat(qty_produce_kg_`crop' yield_`crop'), punc( "()")

matrix prel[rownumb(prel,"`s'"), colnumb(prel,"`d'")]= r(mean)

*matrix prel(`row', `col'+ 2) = `cellval'

local row = `A' + 1

}

local row = `A' + 1

}

}


Any help would be appreciated

[Nick Cox]

Thanks to William Lisowski and John Mullahy for the cute solution: it's just the product of two matrices. On a different note, is there a good reason why runningsum() should not be generalised to work columnwise on a matrix with any number of columns?

[John Mullahy]

y=lowertriangle(J(rows(x),rows(x),1))

[William Lisowski]

Here's a start at a non-looping solution. Creating y as a function of the number of rows in x is left as an exercise for the reader.

: x = (1,2,3\4,5,6\7,8,9)

: x
       1   2   3
    +-------------+
  1 |  1   2   3  |
  2 |  4   5   6  |
  3 |  7   8   9  |
    +-------------+

: y = (1,0,0\1,1,0\1,1,1)

: y*x
        1    2    3
    +----------------+
  1 |   1    2    3  |
  2 |   5    7    9  |
  3 |  12   15   18  |
    +----------------+

[Nick Cox]

This works, and I too would like to see a better way.

: x = (1,2,3\4,5,6\7,8,9)

: x
       1   2   3
    +-------------+
  1 |  1   2   3  |
  2 |  4   5   6  |
  3 |  7   8   9  |
    +-------------+

: wanted = J(rows(x), 0, .)

: for(j = 1; j <= cols(x); j++) {
> wanted = wanted , runningsum(x[, j])
> }

: wanted
        1    2    3
    +----------------+
  1 |   1    2    3  |
  2 |   5    7    9  |
  3 |  12   15   18  |
    +----------------+