Tuesday, April 24, 2012

Matrix of Stata

Matrix of Stata
  • Define a -matrix- or -mat- in short: matrix matrixname = matrix_expression.
    • matrix define mymat = (1,2\3,4)
    • Or, matrix mymat = (1,2\3,4)
    • mat list mymat
  • Print a matrix: 
    • matrix list matrix_name
  • -matlist-: Display a matrix and control its format: 
    • matlist r(table)'*100, tw(20) format(%8.2f)
  • Matrix rename: .matrix rename oldname newname
  • Rename col/row names: 
    • matrix colnames A = names
    • matrix rownames A = names
  • Convert variables to matrix and vice versa: 
    • Matrix addition & subtraction:
      • mat matrix_a + matrix_b
      • mat matrix_a - matrix_b
    • Matrix multiplication:
      • mat matrix_a * matrix_b
    • Transpose of a matrix
      • mat matrix_t = matrix_a'
    • Inverse of a matrix, matrix function are defined by using parentheses ():
      • mat matrix_inv = inv(matrix_a)
    • A\B adds row of B after the rows of A; A,B adds columns B to matrix A. For example:
      • mat RTab=r(table)'
      • mat R1=RTab[1...,"b"]
      • mat R2=RTab[1...,"ll".."ul"]
      • mat RDisp=R1,R2
    • The usual way to obtain matrices after a command that produces matrices is simply to refer to the returned matrix in the standard way. (source: Stata manual) Or 'get()' matrix function obtains a copy of an internal Stata system matrix.
      • For instance all estimation commands return
        • e(b) coefficient vector
        • e(V) variance-covariance matrix of the estimates (VCE)
      • And these matrices can be referenced directly. For examples:
        • matrix list e(b)
        • mattrix myE = e(b)*100
        • matrix myV = e(V)*100
        • matrix myVget=get(_b)
      • Then, you can use
        • "ereturn post myE myV" to Change coefficient vector and variance-covariance matrix.
        • And "ereturn display" to create and display a coefficient table: "r(table)" that have been previously posted using "ereturn post" or "repost". "r(table)" is a matrix containing all the data displayed in the coefficient table (including b, se, t, pvalue, ll, ul, df, crit, and eform).
    • Reset row/column names of matrix
      • mat A = (1,2,3\ 4,5,6\ 7,8,9)
      • matrix rownames A = myrow1 myrow2 myrow3
      • mat colnames A = mycol1 mycol2 mycol3
      • refix the column names of A with equation names eq1, eq2, and eq3: mat coleq A = eq1 eq2 eq3
    • Matrix utilities:
      • matrix dir - List the currently defined matrices
      • matrix list - Display the contents of a matrix
      • matrix rename - Rename a matrix
      • matrix drop - Drop a matrix
    • Matrix Subscripting:
      • Referring to a matrix's element/cell using numbers of row and column or using the names of row and column or using mixed ways:
        • disp matrix_name [r,c]
        • matrix A = A/A[1,1]
        • gen newvar = oldvar/A[2,2]
        • mat B = V["price","price"]
        • gen sdif = dif/sqrt(V["price","price"])
      • Create a matrix B having the first through fifth rows and first through fifth columns of A, using two periods:
        • matrix B = A[1..5,1..5]
      • Create a matrix B having the all rows and first column of A, use three periods: 
        • matrix B = A[1...,1]
        • mat B=A[1...,"b"]
      • Define an element of a maxtrix:
        • mat A[1,2] = sqrt(2)
    • Abstract row and column names from a matrix (-rownames- and -colnames-: Macro extended function related to matrices). Also, we can use -rowfullnames matname-, and -colfullnames matname-
      • sysuse auto,clear
      • regress mpg price weight length
      • local rn :rownames r(table)
      • local cn :colnames r(table)
      • disp "`w1'" // price
      • disp "`:word 1 of `:colnames r(table)''" // price
      • disp "rownames = `rn'" _n "columnnames = `cn'"
      • disp `:word count `cn'' // 4
      • disp "`:word count `:colnames r(table)''" // 4
      • disp "`:colsof r(table)'" // 4
      • disp "`:rowsof r(table)'" // 9
      • forval i = 1/`:word count `cn'' {
      •   local varn: word `i' of `cn'
      •   disp "`varn'"
      •   }
    • Find out number of row and column (-colsof(matname)- and -rowsof(matname)-): Matrix functions
      • disp "number of row = " rowsof(r(table)) 
      • disp "number of column = " colsof(r(table)) 
    • How to do element-by-element operation
      • mata, the new matrix language of Stata (since version 9), can do element-by-element operation easily. (read more here)
      • Colon operations perform element-by-element operations: addition(:+), subtraction(:-), multiplication(:*), division(:/), power(:^), equality(:==), inequality(:!=), greater than(:>), greater than or equal to(:>=), less than(:<), less than or equal to(:<=), and(:&), or(:|)
      • If you apply the exponential function on each element:
        • . mata
        • : x=(0,1\2,3)
        • : exp(x) // x = (1,2.7\7.4,20.1)
        • : exp((0,1\2,3)) // (1,2.7\7.4,20.1)
        • : x:/2 // x = (0,0.5\1, 1.5)
        • : end
      • mata can use a Stata matrix
        • matrix A=r(table)'
        • mata: exp(st_matrix("A")) // display only
        • mata: st_matrix("se",sqrt(diagonal(st_matrix("V")))) // create a matrix of standard deviation
        • mata: st_matrix("expA", exp(st_matrix("A"))) // create a new matrix, expA
        • matrix eformtab=expA[1..., "b"], expA[1...,"ll".."ul"], expA[1...,"se"]
        • matlist eformtab*100, tw(30) format(%8.1f)
    • More about mata
      • create row vector: row=(1,2,3) or row=(1..3)
      • create column vector: col=(1\2\3) or col=(1::3)
      • transpose a matrix: m2=(0,1\2,3)'
      • scalar functions will operate on elements: sqrtrow=sqrt(row)
      • SSCC: an introduction to Mata
    Labels: ,

    No comments:

    Post a Comment

    Subscribe to: Post Comments (Atom)