predinterval - missing values in output

[Richard Armour]

Hello all,

I am completing a meta-analysis of proportions of patients experiencing x outcome. As recommended by Barker et al. I am planning to report predictive intervals as well as confidence intervals in results with high I2 values.

I used command

 meta summarize, proportion predinterval

which produced the following:

  Effect-size label: Freeman–Tukey's p
        Effect size: _meta_es
          Std. err.: _meta_se

Meta-analysis summary                     Number of studies =      2
Random-effects model                      Heterogeneity:
Method: REML                                          tau2 =  0.0533
                                                    I2 (%) =   69.76
                                                        H2 =    3.31

--------------------------------------------------------------------
            Study |     Proportion    [95% conf. interval]  % weight
------------------+-------------------------------------------------
          Study 1 |          0.283       0.161       0.423     51.04
          Study 2 |          0.475       0.321       0.631     48.96
------------------+-------------------------------------------------
 invftukey(theta) |          0.374       0.198       0.569
--------------------------------------------------------------------
95% prediction interval for invftukey(theta): [     .,     .]

Test of theta = 0: z = 5.99                      Prob > |z| = 0.0000
Test of homogeneity: Q = chi2(1) = 3.31            Prob > Q = 0.0690

I appreciate that perhaps the 95% prediction interval is not listed because it is so vanishgly small as to not be relevant, however, for my own understanding I would appreciate any advice on determining the actual predictive interval here.

Thank you,
Richard