Re: st: Inconsistent Result of generating variables

[Richard Williams <Richard.A.Williams.5@nd.edu>]

At 05:12 PM 4/9/2004 -0400, Jitian Sheu wrote:
> Hello, listers:
>
> I am not sure whether it is a bug in Stata?
>
> Before going on, I will like to post my code first, then that would be
> easier to explain:
>
> local b01=0.02
>
> local b11=0.3
>
> local b21=-0.25
>
> gen tmptau1=emusq1+`b01'^2+(`b11'^2)*(x1^2)+(`b21'^2)*(x1^4)-2*`b01'*emu1-2*
> /*
Those are complicated expressions!  But the problem, I believe, is that b21 
is a negative number.  There is a FAQ on this somewhere, but I don't recall 
where it is.  Basically, in the b21 part of the expression above, what you 
probably want is

-.25^2 * x1^4 = (-.25)^2 * x1 ^ 4 = .0625 * x1 ^ 4

but what Stata is giving you is

-(.25^2) * x1^4 = -.0625 * x1^4

i.e. you are subtracting when you want to add, hence your first measure 
always winds up smaller than your second (even when they appear the same, 
they are different several decimal places down).

To fix this, in the above line, put another set of parentheses around 
'b21', i.e. make it


> gen tmptau1=emusq1+`b01'^2+(`b11'^2)*(x1^2)+((`b21')^2)*(x1^4)-2*`b01'*emu1-2*
When I do that, I get

. list tmptau1 tmptau1_alt

     +---------------------+
     |  tmptau1   tmptau~t |
     |---------------------|
  1. | .8692592   .8692592 |
  2. | .2181771   .2181771 |
  3. | .2075188   .2075188 |
  4. | .1309747   .1309747 |
  5. | .6345458   .6345458 |
     |---------------------|
  6. | .4080512   .4080512 |
  7. | .2966675   .2966675 |
  8. | .3682937   .3682937 |
  9. | .0604092   .0604092 |
 10. | .9241688   .9241688 |
     +---------------------+

Incidentally, Stata is doing it right; exponentiation gets done before 
subtraction.  e.g. if I punch in 3 - 4^2 on my calculator, I get -13, not 
19. If I punch in 3 + (-4)^2, I get 19. But, it is easy to get confused about!

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/