NOTE: RoBMA — R vs JASP Implementation
=======================================

Paper: Publication bias and p-hacking in the effect of COVID-19 on learning
Authors: Luskova, Buliskeria, Elminejad, Havranek, Irsova, Jurajda, Kapicka

Background
----------
Robust Bayesian Meta-Analysis (RoBMA) can be run either in R (via the RoBMA
package) or in JASP (a free, open-source statistical software with a graphical
interface). Both implementations use the same underlying methodology but may
produce slightly different results due to MCMC sampling variability and minor
differences in default priors across versions.

Results reported in the paper
------------------------------
The RoBMA estimate reported in the paper (Table 3, Panel B) was obtained in
JASP:

  Mean effect size (mu):  -0.118
  95% credible interval:  [-0.135, -0.094]
  Heterogeneity (tau):     0.115
  Inclusion BF for bias:   1.682

R replication result
--------------------
Running RoBMA in R (version 3.4.0, seed = 1, effect_direction = "negative")
produces:

  Mean effect size (mu):  -0.120
  95% credible interval:  [-0.136, -0.100]
  Heterogeneity (tau):     0.115

The small difference (-0.118 vs -0.120) reflects MCMC sampling variability
and minor differences in default priors between JASP and R implementations.
Both estimates are substantively identical and support the same conclusions.

Running RoBMA in R
------------------
RoBMA in R requires JAGS to be installed separately:

  1. Download JAGS 4.x from: https://sourceforge.net/projects/mcmc-jags/
  2. Install the .pkg file (macOS) or .exe file (Windows)
  3. Verify in R: rjags::jags.version()
  4. Run the RoBMA block in code/02_bias_correction.R

Note: RoBMA is slow to run (approximately 10-30 minutes).

Running RoBMA in JASP
---------------------
If the R implementation fails or produces unexpected results:

  1. Download JASP from: https://jasp-stats.org/
  2. Open JASP and load data/learning_loss_data.xlsx
  3. Go to Meta-Analysis -> Robust Bayesian Meta-Analysis
  4. Set:
       - Effect size: es
       - Standard error: se
       - Effect direction: negative
       - Seed: 1
  5. The model-averaged estimates correspond to Table 3, Panel B (RoBMA row)

References
----------
Bartos, F., Maier, M., Wagenmakers, E.-J., Doucouliagos, H. and Stanley, T.
  (2023). Robust Bayesian meta-analysis: Model-averaging across complementary
  publication bias adjustment methods. Research Synthesis Methods, 14(1), 99-116.

Maier, M., Bartos, F. and Wagenmakers, E.-J. (2023). Robust Bayesian
  meta-analysis: Addressing publication bias with model-averaging.
  Psychological Methods, 28(1), 107.
