Assigning more weight to a category of a variable

[Giorgio Piccitto]

Dear all,

I would like to know your opinion about the following issue: I need to create (starting from my original dataset) a series of 'fictious datasets' in which I simulate another scenario changing the distribution of a given characteristics, and then calculating (for each 'fictious dataset') the correlation matrix to see if it consistent with the one I have in the original dataset

For example, In my date I have the 20% of female and 80% of male, and I want to obtain that they are 50% and 50%.

I proceeded as follows:
- calculating the weight (50/80 for male, 50/20 for female)
- creating a new variable that assigns the value 50/80 to male and 50/20 to female
- running the code "corr soldi donna [fweight=weight]", with weight being the variable which contains the weights for male and female

Anyway, Stata gives me an error stating "may not use noninteger frequency weights", where I am wrong? Is not that the kind of weight I need to do this procedure?

Thanks a lot, best, G.

[Carlo Lazzaro]

Giorgio:
elaborating a bit on your example, why not using -regress-?:

. set obs 10
number of observations (_N) was 0, now 10

. g id=_n

. g female=1 in 1/2
(8 missing values generated)

. replace female=0 if female==.
(8 real changes made)

. label define female 0 "male" 1 "female"

. label values female female

. g money=runiform()*1000

. g iw=50/20 if female==1
(8 missing values generated)

. replace iw=50/80 if female==0
(8 real changes made)

. regress money i.female [iw=iw]

      Source |       SS           df       MS      Number of obs   =        10
-------------+----------------------------------   F(1, 8)         =      0.35
       Model |  21828.1041         1  21828.1041   Prob > F        =    0.5727
    Residual |   504835.68         8    63104.46   R-squared       =    0.0414
-------------+----------------------------------   Adj R-squared   =   -0.0784
       Total |  526663.784         9  58518.1982   Root MSE        =    251.21

------------------------------------------------------------------------------
       money |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      female |
     female  |  -93.44111   158.8766    -0.59   0.573    -459.8113    272.9291
       _cons |   401.3198   112.3427     3.57   0.007      142.257    660.3827
------------------------------------------------------------------------------