Skip to main content

Table 6 Summary results of meta-analysis using subsets of fish abundance effect sizes for interannual Before/After studies, testing the influence of the given moderator variable

From: How do changes in flow magnitude due to hydropower operations affect fish abundance and biomass in temperate regions? A systematic review

Moderator

k

Q statistic (p-value)

QM (p-value)

QE (p-value)

Waterbody type

112

N/A

  

Dam size

112

N/A

  

Hydropower operational regime

    

 Unmoderated model

112

421.12 (p < 0.0001)

–

–

 Operational regime

112

–

1.35 (p = 0.717)

412.00 (p < 0.0001)

Direction of flow magnitude alteration

    

 Unmoderated model

112

421.12 (p < 0.0001)

–

–

 Flow alteration

112

–

9.07 (p = 0.028)

412.05 (p < 0.0001)

Alterations to other flow components

    

 Unmoderated model

112

421.12 (p < 0.0001)

–

–

 Other components

112

–

29.42 (p < 0.0001)

391.70 (p < 0.0001)

Sampling method

    

 Unmoderated model

112

421.12 (p < 0.0001)

–

–

 Sampling technique

112

–

19.83 (p = 0.0002)

401.29 (p < 0.0001)

Sampling season

    

 Unmoderated model

112

421.12 (p < 0.0001)

–

–

 Sampling season

112

–

10.11 (p = 0.039)

411.01 (p < 0.0001)

Type of comparator (temporal)

    

 Unmoderated model

112

421.12 (p < 0.0001)

–

–

 Temporal comparison

112

–

1.63 (p = 0.203)

420.92 (p < 0.0001)

Time since intervention

    

 Unmoderated model

111

420.78 (p < 0.0001)

–

–

 Time since intervention

111

–

1.96 (p = 0.376)

418.83 (p < 0.0001)

Monitoring duration (with outliers†)

    

 Unmoderated model

112

421.12 (p < 0.0001)

–

–

 Monitoring duration

112

–

1.32 (p = 0.252)

417.43 (p < 0.0001)

Life stage

    

 Unmoderated model

108

415.73 (p < 0.0001)

–

–

 Life stage

108

–

35.36 (p < 0.0001)

380.12 (p < 0.0001)

  1. Bold indicates statistical significance (p < 0.05)
  2. k: number of effect sizes; Q statistic: value of homogeneity test; QM: omnibus test statistic of moderators; QE: unexplained heterogeneity
  3. Unmoderated model: random-effects model; *Significance at p < 0.1. †Two extreme effect sizes were removed to improve model fit, but removal had little impact on results.