Re: st: Significance values using -prais-

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

At 03:28 AM 1/7/2004 +0000, Clive Nicholas wrote:
> All,
>
> First of all, happy new year to all of you!
>
> I wondered if anybody has noticed that, when using -prais-, sometimes the
> output model contains t-values that clearly exceed 1.96, but yet show up
> as statistically insignificant, even at p<.05 (I don't regard p<.10 as a
> particularly useful benchmark for significance testing). Here's an
> example, with <enp> the 'offending' variable here:
I've never used this routine, but it looks like the degrees of freedom for 
the t-tests are 20, right?  Which means that the critical values for a 
given alpha level will be a bit larger than they would be for a 
z-score.  For a T with 20 d.f. the critical two-tailed value for the .05 
level of significance is 2.086.

You can double-check the reported sig. levels this way:

. tablesq T 20 4

t(20) = 4
       Pr(T <= 4)   = 0.9996
       Pr(T >= 4)   = 0.0004
       Pr(|T| >= 4) = 0.0007

. tablesq T 20 1.76

t(20) = 1.76
       Pr(T <= 1.76)   = 0.9532
       Pr(T >= 1.76)   = 0.0468
       Pr(|T| >= 1.76) = 0.0937

. tablesq T 20 2.04

t(20) = 2.04
       Pr(T <= 2.04)   = 0.9726
       Pr(T >= 2.04)   = 0.0274
       Pr(|T| >= 2.04) = 0.0548

. tablesq T 20 7.25

t(20) = 7.25
       Pr(T <= 7.25)   = 1.0000
       Pr(T >= 7.25)   = 0.0000
       Pr(|T| >= 7.25) = 0.0000

. tablesq T 20 1.67

t(20) = 1.67
       Pr(T <= 1.67)   = 0.9448
       Pr(T >= 1.67)   = 0.0552
       Pr(|T| >= 1.67) = 0.1105

-------------------------------------------
Richard Williams, Associate Professor
OFFICE: (574)631-6668, (574)631-6463
FAX:    (574)288-4373
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW (personal):    http://www.nd.edu/~rwilliam
WWW (department):    http://www.nd.edu/~soc

*
*   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/