Table 1 Summary of Bayesian model-averaged meta-analysis of grammatical impairment in schizophrenia spectrum disorders

From: Relationship between grammar and schizophrenia: a systematic review and meta-analysis

Syntactic domain

n

N PwSz, Controls

Cohen’s d (BMA:95% CrI)

BF10 for H1

Heterogeneity Tau (95% CrI)

BFrf for RE

Significant Moderator effect (s.e.)

lnCVR (BMA:95% CrI)

ROBMA BF10 for publication bias & mean difference

Syntactic comprehension

16

530/400

1.01 [0.85, 1.19]

220 ×105 Extreme

0.18 [0.04, 0.42]

1.08 Weak

None

0.41 [0.11, 0.71]

Weak publ. bias (1.53) Extreme effect (244)

Error detection

6

170/134

0.91 [0.60, 1.19]

303 Extreme

0.21 [0.04, 0.61]

1.01 Weak

None

0.25 [−0.52, 0.96]

No publ. bias (0.72) Strong effect (23.0)

Production length

17

646/614

0.84 [0.63, 1.04]

460 ×103 Extreme

0.31 [0.15, 0.53]

313.75 Extreme

None

0.13 [0.00, 0.25]

Weak publ. bias (2.78) Strong effect (7.78)

Phrasal complexity

16

489/325

0.63 [0.46, 0.81]

335 ×102 Extreme

0.20 [0.05, 0.45]

1.78 Weak

None

0.29 [0.03, 0.56]

Weak publ. bias (1.41) Moderate effect (3.52)

Production integrity

11

368/303

0.73 [0.49, 0.99]

1258 Extreme

0.27 [0.06, 0.60]

4.67 Strong

Quality 0.31 (0.12) P = 0.009

0.12 [−0.08, 0.32]

No publ. bias (0.96) Strong effect (9.95)

Global complexity

13

335/246

0.65 [0.39, 0.92]

362 Extreme

0.35 [0.14, 0.65]

54.86 Very Strong

Agea 0.04 (0.01) p = 0.008

0.22 [−0.01, 0.43]

Weak publ. bias (1.06) Moderate effect (4.42)

  1. ROBMA Robust Bayesian meta-analysis, s.e. Standard Error, n number of studies, N sample size based on unique participant counts, CrI Credible Intervals, BFrf Bayes Factor for random effects over fixed effects, BF10 Bayes Factor for evidence for the presence of expected group differences over the null hypothesis of no difference, BMA Bayesian Model Average, lnCVR natural log of the coefficient of variation ratio for patients vs. controls, RE random effects. Note that due to an editorial instruction to avoid the term patients we use the phrase PwSz people with schizophrenia; but all the individual studies included in this meta-analysis refer to ‘patients’ i.e., people who seek clinical help for their symptoms of psychosis.
  2. aEstimated from n = 12; higher deficits in samples with higher mean age.