COMPGROUPS.NET | Search | Post Question | Groups | Stream | About | Register

### Proc Surveyfreq standard errors. Explanation?

• Follow

```Hi all,
Just noticed that when my variables have 2 levels (eg Yes/No) the standard
errors for 'Yes' and 'No' are the same.  If I use a BY statement then the
SEs are different across the BY groups but still the same within them. The
SEs change are different if I have more than 2 levels (eg Yes/No/Dont know).

HSAE    F_q48ace        WgtFreq Percent StdErr
LNS HS  A_Yes   95102.37        38.59   2.215913965
LNS HS  B_No    150867.71       61.21   2.2188699
LNS HS  C_Dont know     492.22  0.20    0.199704623
LNS HS  Total   246462.30       100.00
NB HSA  A_Yes   79506.81        42.61   2.731449697
NB HSA  B_No    107085.11       57.39   2.731449697
NB HSA  Total   186591.92       100.00

How come?
L_B

"The beggar laughs in the face of the thief"

_________________________________________________________________
http://a.ninemsn.com.au/b.aspx?URL=http%3A%2F%2Fninemsn%2Erealestate%2Ecom%2Eau%2Fcgi%2Dbin%2Frsearch%3Fa%3Dbhp%26t%3Dres%26cu%3DMSN&_t=758874163&_r=HM_EndText_Nov06&_m=EXT
```
 0
Reply laughing_beggar (62) 11/27/2006 1:21:47 AM

```Recall that the variance of a binomial proportion is square root
[PQ/n]. This is the simplest case with 2 levels, when Q = 1-P. Thus the
Variance of 'Yes' responses will neccesarily equal Variance of 'No'.

Of course under complex surveys http://tinyurl.com/ux8p9 variance
computations are a lot more complicated but the 'fundamentals' will
always be the same.

```
 0
Reply tanwanzang (263) 11/27/2006 3:55:53 PM

1 Replies
56 Views

Similiar Articles:

7/6/2012 4:35:19 PM