Re: st: RE: How to calculate P values and I2 from estore after metareg?

[Roger Harbord <rogerharbord@bigfoot.com>]

Many thanks Tom.

Just to clarify that when e(df_r) is set, it's the same as e(df_Q). It's
deliberately not set if you specify the "z" option to turn off the
Knapp-Hartung variance adjustment and associated use of t-tests instead
of  z-tests for the coefficients,  as it's setting e(df_r) that makes
-ereturn display- report t and F rather than z and chi2.

STB-42 concerned the original version of -metareg-, written by Stephen
Sharp. I've finally got around to drafting an article on the update to
-metareg- and hope to submit it to Stata Journal before too long.

Roger.

Steichen, Thomas J. wrote:
> Geoff. Livesey writes (in part):
> ___________________________________________________________________
>
> I am wanting output the results of metareg to an excel file.
>
> I can calculate certain P values as follows:
>
> gen  slope = _b[X]
> gen  slope_se =_se[X]
> gen  slope_t = slope/slope_se
> gen  slope_p = 2* ttail(df_r,abs(slope_t))
>
> gen  const =_b[_cons]
> gen  const_se =_se[_cons]
> gen  const_t = const/const_se
> gen  const_p = 2*ttail(df_r,abs(const_t))
>
> gen  Q_p = chi2tail(df_r,e(Q))
>
> However, it is not clear to me how to calculate:
>
> I2 - I can get this approximately via an inelegant route,
> however I should like to know more and how to obtain this
> information directly from that stored in e. How is this calculated? The
> precise value given in the metareg immediately prior to the above
> ereturn information is 0.695
>
> P_Tau2 - Likewise, how is P for Tau2 derived from
> information in the e store?
> The precise value given in the metareg immediately prior to the above
> ereturn information is 0.1836
>
> Q_p - The value of P for Q I calculate is given by the formula above,
> that is it works with this example but is the procedure correct? The
> precise value given by metareg immediately prior to the above ereturn
> information is 0.020.
> _____________________________________________________________________
>
> The code in metareg shows:
>
> For I2:
>
>    scalar i2 = max( ( e(Q) - e(df_Q) ) / e(Q), 0)
>
> For P_Tau2:
>
>    scalar pval =  chi2tail(1, e(chi2_c))*0.5
>    if e(chi2_c) <=0 {
> 	scalar `pval'= 1
>    }
>
> Which can be written in one line as:
>
>    scalar pval = cond(e(chi2_c) <= 0, 1, chi2tail(1, e(chi2_c))*0.5)
>
> For Q_p:
>
>    scalar pval = chi2tail(e(df_Q), e(Q))
>
> Which indicates that you are using the wrong return value for df.
>
> All the needed values are in the ereturn results.
>
> Tom
*
*   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/