----------------------------------------------------------------------------------------------------------------------- name: log: H:\missing\meta\Egger chapter\final\practical.log log type: text opened on: 3 Mar 2015, 12:38:54 Log file practical.log opened on 3 Mar 2015 at 12:38:54 . // Stata practical for missing data in meta-analysis . // IW, 3mar2015 . // This do-file does the basics without trying to beautify the graphs . . // Part 1: Haloperidol data . . * Q1. Start by drawing a forest plot and doing a simple meta-analysis (ACA) using . use haloperidol, clear (Haloperidol meta-analysis (Joy et al 2006) + missing data (Higgins et al 2008)) . metan rh fh rp fp, rr fixedi label(namevar=author) Study | RR [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.417 0.891 2.252 18.86 Beasley (1996) | 1.049 0.732 1.504 31.22 Bechelli (1983) | 6.207 1.520 25.353 2.05 Borison (1992) | 7.000 0.400 122.442 0.49 Chouinard (1993) | 3.492 1.113 10.955 3.10 Durost (1964) | 8.684 1.258 59.946 1.09 Garry (1962) | 1.750 0.585 5.238 3.37 Howard (1974) | 2.039 0.670 6.208 3.27 Marder (1994) | 1.357 0.747 2.466 11.37 Nishikawa (1982) | 3.000 0.137 65.903 0.42 Nishikawa (1984) | 9.200 0.581 145.759 0.53 Reschke (1974) | 3.793 1.058 13.604 2.48 Selman (1976) | 1.484 0.936 2.352 19.11 Serafetinides (1972) | 8.400 0.496 142.271 0.51 Simpson (1967) | 2.353 0.127 43.529 0.48 Spencer (1992) | 11.000 1.671 72.396 1.14 Vichaiya (1971) | 19.000 1.157 311.957 0.52 ---------------------+--------------------------------------------------- I-V pooled RR | 1.567 1.281 1.916 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 27.29 (d.f. = 16) p = 0.038 I-squared (variation in RR attributable to heterogeneity) = 41.4% Test of RR=1 : z= 4.37 p = 0.000 . . * Q2. Explore the missing data. . gen missfrach = mh/(rh+fh+mh) . gen missfracp = mp/(rp+fp+mp) . gen id = substr(author,1,2) // two-letter abbreviation to label the next graph . scatter missfrac*, mlab(id) . . * Q3. Download metamiss if necessary . *net from http://www.mrc-bsu.cam.ac.uk/IW_Stata/meta/ . *net install metamiss . . * Q4. Replicate the metan analysis using . metamiss rh fh mh rp fp mp, rr fixed id(author) aca ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Available cases analysis ******** ******************************************************************* Measure: RR. Zero cells detected: adding 1/2 to 6 studies. (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.417 0.891 2.252 18.86 Beasley (1996) | 1.049 0.732 1.504 31.22 Bechelli (1983) | 6.207 1.520 25.353 2.05 Borison (1992) | 7.000 0.400 122.442 0.49 Chouinard (1993) | 3.492 1.113 10.955 3.10 Durost (1964) | 8.684 1.258 59.946 1.09 Garry (1962) | 1.750 0.585 5.238 3.37 Howard (1974) | 2.039 0.670 6.208 3.27 Marder (1994) | 1.357 0.747 2.466 11.37 Nishikawa (1982) | 3.000 0.137 65.903 0.42 Nishikawa (1984) | 9.200 0.581 145.759 0.53 Reschke (1974) | 3.793 1.058 13.604 2.48 Selman (1976) | 1.484 0.936 2.352 19.11 Serafetinides (1972) | 8.400 0.496 142.271 0.51 Simpson (1967) | 2.353 0.127 43.529 0.48 Spencer (1992) | 11.000 1.671 72.396 1.14 Vichaiya (1971) | 19.000 1.157 311.957 0.52 ---------------------+--------------------------------------------------- I-V pooled ES | 1.567 1.281 1.916 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 27.29 (d.f. = 16) p = 0.038 I-squared (variation in ES attributable to heterogeneity) = 41.4% Test of ES=1 : z= 4.37 p = 0.000 . . * Q5. Perform an ICA-0 analysis by replacing aca by ica0 above. . * Similarly you can use the ica1, icapc and icap options . metamiss rh fh mh rp fp mp, rr fixed id(author) ica0 ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Simple imputation ******** ******************************************************************* Measure: RR. Method: ICA-0 (impute zeros). Weighting scheme: w4. Zero cells detected: adding 1/2 to 6 studies. (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.362 0.854 2.172 24.38 Beasley (1996) | 1.429 0.901 2.266 25.01 Bechelli (1983) | 6.200 1.513 25.402 2.67 Borison (1992) | 7.000 0.400 122.442 0.65 Chouinard (1993) | 3.492 1.113 10.955 4.06 Durost (1964) | 8.684 1.258 59.946 1.42 Garry (1962) | 1.750 0.582 5.266 4.38 Howard (1974) | 2.039 0.670 6.208 4.29 Marder (1994) | 1.357 0.745 2.473 14.75 Nishikawa (1982) | 3.000 0.137 65.903 0.56 Nishikawa (1984) | 8.474 0.534 134.463 0.70 Reschke (1974) | 3.793 1.058 13.604 3.26 Selman (1976) | 2.429 1.189 4.960 10.42 Serafetinides (1972) | 9.000 0.530 152.927 0.66 Simpson (1967) | 2.647 0.142 49.419 0.62 Spencer (1992) | 11.000 1.671 72.396 1.50 Vichaiya (1971) | 19.000 1.156 312.417 0.68 ---------------------+--------------------------------------------------- I-V pooled ES | 1.898 1.507 2.390 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 21.56 (d.f. = 16) p = 0.158 I-squared (variation in ES attributable to heterogeneity) = 25.8% Test of ES=1 : z= 5.45 p = 0.000 . metamiss rh fh mh rp fp mp, rr fixed id(author) ica1 ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Simple imputation ******** ******************************************************************* Measure: RR. Method: ICA-1 (impute ones). Weighting scheme: w4. Zero cells detected: adding 1/2 to 3 studies. (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.471 0.934 2.317 5.95 Beasley (1996) | 0.931 0.773 1.120 35.81 Bechelli (1983) | 4.478 1.417 14.151 0.93 Borison (1992) | 7.000 0.400 122.442 0.15 Chouinard (1993) | 3.492 1.113 10.955 0.94 Durost (1964) | 8.684 1.258 59.946 0.33 Garry (1962) | 1.600 0.603 4.247 1.29 Howard (1974) | 2.039 0.670 6.208 0.99 Marder (1994) | 1.313 0.754 2.283 4.01 Nishikawa (1982) | 3.000 0.137 65.903 0.13 Nishikawa (1984) | 10.684 0.682 167.429 0.16 Reschke (1974) | 3.793 1.058 13.604 0.75 Selman (1976) | 1.120 0.953 1.316 47.38 Serafetinides (1972) | 4.000 0.509 31.456 0.29 Simpson (1967) | 1.000 0.106 9.444 0.24 Spencer (1992) | 11.000 1.671 72.396 0.35 Vichaiya (1971) | 10.000 1.364 73.328 0.31 ---------------------+--------------------------------------------------- I-V pooled ES | 1.156 1.035 1.292 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 40.34 (d.f. = 16) p = 0.001 I-squared (variation in ES attributable to heterogeneity) = 60.3% Test of ES=1 : z= 2.57 p = 0.010 . metamiss rh fh mh rp fp mp, rr fixed id(author) icapc ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Simple imputation ******** ******************************************************************* Measure: RR. Method: ICA-pC (impute control group risk). Weighting scheme: w4. Zero cells detected: adding 1/2 to 6 studies. (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.401 0.880 2.230 19.89 Beasley (1996) | 1.033 0.718 1.486 32.56 Bechelli (1983) | 6.033 1.475 24.682 2.17 Borison (1992) | 7.000 0.400 122.442 0.53 Chouinard (1993) | 3.492 1.113 10.955 3.29 Durost (1964) | 8.684 1.258 59.946 1.15 Garry (1962) | 1.721 0.574 5.161 3.57 Howard (1974) | 2.039 0.670 6.208 3.47 Marder (1994) | 1.346 0.741 2.448 12.05 Nishikawa (1982) | 3.000 0.137 65.903 0.45 Nishikawa (1984) | 8.553 0.539 135.713 0.56 Reschke (1974) | 3.793 1.058 13.604 2.64 Selman (1976) | 1.300 0.759 2.228 14.85 Serafetinides (1972) | 8.400 0.496 142.271 0.54 Simpson (1967) | 2.353 0.127 43.529 0.51 Spencer (1992) | 11.000 1.671 72.396 1.21 Vichaiya (1971) | 18.419 1.121 302.646 0.55 ---------------------+--------------------------------------------------- I-V pooled ES | 1.530 1.243 1.883 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 27.61 (d.f. = 16) p = 0.035 I-squared (variation in ES attributable to heterogeneity) = 42.0% Test of ES=1 : z= 4.02 p = 0.000 . metamiss rh fh mh rp fp mp, rr fixed id(author) icap ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Simple imputation ******** ******************************************************************* Measure: RR. Method: ICA-p (impute group-specific risk). Weighting scheme: w4. Zero cells detected: adding 1/2 to 6 studies. (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.417 0.891 2.252 18.86 Beasley (1996) | 1.049 0.732 1.504 31.22 Bechelli (1983) | 6.207 1.520 25.353 2.05 Borison (1992) | 7.000 0.400 122.442 0.49 Chouinard (1993) | 3.492 1.113 10.955 3.10 Durost (1964) | 8.684 1.258 59.946 1.09 Garry (1962) | 1.750 0.585 5.238 3.37 Howard (1974) | 2.039 0.670 6.208 3.27 Marder (1994) | 1.357 0.747 2.466 11.37 Nishikawa (1982) | 3.000 0.137 65.903 0.42 Nishikawa (1984) | 9.200 0.581 145.759 0.53 Reschke (1974) | 3.793 1.058 13.604 2.48 Selman (1976) | 1.484 0.936 2.352 19.11 Serafetinides (1972) | 8.400 0.496 142.271 0.51 Simpson (1967) | 2.353 0.127 43.529 0.48 Spencer (1992) | 11.000 1.671 72.396 1.14 Vichaiya (1971) | 19.000 1.157 311.957 0.52 ---------------------+--------------------------------------------------- I-V pooled ES | 1.567 1.281 1.916 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 27.29 (d.f. = 16) p = 0.038 I-squared (variation in ES attributable to heterogeneity) = 41.4% Test of ES=1 : z= 4.37 p = 0.000 . . * Q6. Look at the data on reasons . list author d* +----------------------------------------------------------------------+ | author dfh dsh dch dgh dfp dsp dcp dgp | |----------------------------------------------------------------------| 1. | Arvanitis (1997) 17 0 17 0 30 0 5 0 | 2. | Beasley (1996) 19 0 15 5 32 0 13 1 | 3. | Bechelli (1983) 0 0 0 1 0 0 0 1 | 4. | Borison (1992) 0 0 0 0 0 0 0 0 | 5. | Chouinard (1993) 11 0 2 0 10 0 6 0 | |----------------------------------------------------------------------| 6. | Durost (1964) 0 0 0 0 0 0 0 0 | 7. | Garry (1962) 0 0 1 0 0 0 1 0 | 8. | Howard (1974) 0 0 0 0 0 0 0 0 | 9. | Marder (1994) 25 0 0 13 41 0 0 4 | 10. | Nishikawa (1982) 0 0 0 0 0 0 0 0 | |----------------------------------------------------------------------| 11. | Nishikawa (1984) . . . . 0 0 0 0 | 12. | Reschke (1974) 0 0 0 2 6 0 0 0 | 13. | Selman (1976) 4 0 0 7 8 0 0 10 | 14. | Serafetinides (1972) 0 0 0 0 1 0 0 0 | 15. | Simpson (1967) 0 0 0 0 . . . . | |----------------------------------------------------------------------| 16. | Spencer (1992) 0 0 0 0 0 0 0 0 | 17. | Vichaiya (1971) 0 0 0 1 0 0 0 1 | +----------------------------------------------------------------------+ . . * Q7. Perform the meta-analysis using reasons . metamiss rh fh mh rp fp mp, rr id(author) fixed ica0(dfh dfp) ica1(dsh dsp) icapc(dch dcp) icap(dgh dgp) ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Imputation using reasons ******** ******************************************************************* Measure: RR. Method: ICA-r combining ICA-0 ICA-1 ICA-pC ICA-p. Weighting scheme: w4. Zero cells detected: adding 1/2 to 6 studies. (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.381 0.867 2.201 21.37 Beasley (1996) | 1.349 0.892 2.041 27.10 Bechelli (1983) | 6.207 1.520 25.353 2.34 Borison (1992) | 7.000 0.400 122.442 0.57 Chouinard (1993) | 3.492 1.113 10.955 3.55 Durost (1964) | 8.684 1.258 59.946 1.24 Garry (1962) | 1.721 0.574 5.161 3.85 Howard (1974) | 2.039 0.670 6.208 3.75 Marder (1994) | 1.368 0.751 2.491 12.91 Nishikawa (1982) | 3.000 0.137 65.903 0.49 Nishikawa (1984) | 8.644 0.545 137.115 0.61 Reschke (1974) | 3.793 1.058 13.604 2.85 Selman (1976) | 1.767 1.037 3.010 16.36 Serafetinides (1972) | 9.000 0.530 152.927 0.58 Simpson (1967) | 2.568 0.138 47.821 0.54 Spencer (1992) | 11.000 1.671 72.396 1.31 Vichaiya (1971) | 19.000 1.157 311.957 0.59 ---------------------+--------------------------------------------------- I-V pooled ES | 1.785 1.439 2.214 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 21.86 (d.f. = 16) p = 0.148 I-squared (variation in ES attributable to heterogeneity) = 26.8% Test of ES=1 : z= 5.27 p = 0.000 . . * Q8. Take an IMOR of 0.5 . metamiss rh fh mh rp fp mp, rr id(author) fixed imor(0.5) ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Simple imputation ******** ******************************************************************* Measure: RR. Method: ICA-IMOR (impute using IMORs 0.5 0.5). Weighting scheme: w4. Zero cells detected: adding 1/2 to 6 studies. (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.399 0.878 2.227 22.12 Beasley (1996) | 1.120 0.737 1.700 27.47 Bechelli (1983) | 6.227 1.521 25.488 2.41 Borison (1992) | 7.000 0.400 122.442 0.58 Chouinard (1993) | 3.492 1.113 10.955 3.66 Durost (1964) | 8.684 1.258 59.946 1.28 Garry (1962) | 1.753 0.583 5.265 3.96 Howard (1974) | 2.039 0.670 6.208 3.87 Marder (1994) | 1.358 0.746 2.473 13.34 Nishikawa (1982) | 3.000 0.137 65.903 0.50 Nishikawa (1984) | 8.908 0.562 141.280 0.63 Reschke (1974) | 3.793 1.058 13.604 2.94 Selman (1976) | 1.743 0.973 3.121 14.11 Serafetinides (1972) | 8.684 0.512 147.438 0.60 Simpson (1967) | 2.487 0.134 46.312 0.56 Spencer (1992) | 11.000 1.671 72.396 1.35 Vichaiya (1971) | 19.056 1.159 313.198 0.61 ---------------------+--------------------------------------------------- I-V pooled ES | 1.699 1.365 2.115 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 24.63 (d.f. = 16) p = 0.077 I-squared (variation in ES attributable to heterogeneity) = 35.0% Test of ES=1 : z= 4.75 p = 0.000 . . * Q9. Allow uncertainty about this IMOR . metamiss rh fh mh rp fp mp, rr id(author) fixed logimor(-0.35) sdlogimor(0.35) ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Bayesian analysis using priors ******** ******************************************************************* Measure: RR. Zero cells detected: adding 1/2 to 6 studies. Priors used: Group 1: N(-0.35,0.35^2). Group 2: N(-0.35,0.35^2). Correlation: 0. Method: Gauss-Hermite quadrature (10 integration points). (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.407 0.884 2.240 22.17 Beasley (1996) | 1.084 0.705 1.667 25.89 Bechelli (1983) | 6.217 1.520 25.431 2.41 Borison (1992) | 7.000 0.400 122.442 0.58 Chouinard (1993) | 3.492 1.113 10.955 3.66 Durost (1964) | 8.684 1.258 59.947 1.28 Garry (1962) | 1.751 0.584 5.255 3.97 Howard (1974) | 2.039 0.670 6.208 3.86 Marder (1994) | 1.358 0.746 2.471 13.36 Nishikawa (1982) | 3.000 0.137 65.903 0.50 Nishikawa (1984) | 9.051 0.571 143.520 0.63 Reschke (1974) | 3.793 1.058 13.604 2.94 Selman (1976) | 1.609 0.925 2.800 15.62 Serafetinides (1972) | 8.542 0.503 144.933 0.60 Simpson (1967) | 2.419 0.130 44.981 0.56 Spencer (1992) | 11.000 1.671 72.396 1.35 Vichaiya (1971) | 19.027 1.158 312.638 0.61 ---------------------+--------------------------------------------------- I-V pooled ES | 1.677 1.348 2.088 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 24.95 (d.f. = 16) p = 0.071 I-squared (variation in ES attributable to heterogeneity) = 35.9% Test of ES=1 : z= 4.63 p = 0.000 . . * Q10. Allow the IMORs to be correlated . metamiss rh fh mh rp fp mp, rr id(author) fixed logimor(-0.35) sdlogimor(0.35) corr(0.5) ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Bayesian analysis using priors ******** ******************************************************************* Measure: RR. Zero cells detected: adding 1/2 to 6 studies. Priors used: Group 1: N(-0.35,0.35^2). Group 2: N(-0.35,0.35^2). Correlation: 0.5. Method: Gauss-Hermite quadrature (10 integration points). (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.407 0.884 2.240 21.67 Beasley (1996) | 1.084 0.717 1.639 27.36 Bechelli (1983) | 6.217 1.520 25.429 2.36 Borison (1992) | 7.000 0.400 122.442 0.57 Chouinard (1993) | 3.492 1.113 10.955 3.58 Durost (1964) | 8.684 1.258 59.947 1.25 Garry (1962) | 1.751 0.584 5.255 3.88 Howard (1974) | 2.039 0.670 6.208 3.78 Marder (1994) | 1.358 0.746 2.470 13.07 Nishikawa (1982) | 3.000 0.137 65.903 0.49 Nishikawa (1984) | 9.051 0.571 143.520 0.61 Reschke (1974) | 3.793 1.058 13.604 2.87 Selman (1976) | 1.609 0.928 2.791 15.45 Serafetinides (1972) | 8.542 0.503 144.933 0.58 Simpson (1967) | 2.419 0.130 44.981 0.55 Spencer (1992) | 11.000 1.671 72.396 1.32 Vichaiya (1971) | 19.027 1.158 312.625 0.60 ---------------------+--------------------------------------------------- I-V pooled ES | 1.662 1.339 2.064 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 25.27 (d.f. = 16) p = 0.065 I-squared (variation in ES attributable to heterogeneity) = 36.7% Test of ES=1 : z= 4.60 p = 0.000 . . * Q11. Try larger levels of uncertainty. . metamiss rh fh mh rp fp mp, rr id(author) fixed logimor(-0.35) sdlogimor(3) corr(0.5) ******************************************************************* ******** METAMISS: meta-analysis allowing for missing data ******** ******** Bayesian analysis using priors ******** ******************************************************************* Measure: RR. Zero cells detected: adding 1/2 to 6 studies. Priors used: Group 1: N(-0.35,3^2). Group 2: N(-0.35,3^2). Correlation: 0.5. Method: Gauss-Hermite quadrature (10 integration points). (Calling metan with options: label(namevar=author) fixed eform ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Arvanitis (1997) | 1.412 0.884 2.254 30.61 Beasley (1996) | 1.119 0.531 2.358 12.05 Bechelli (1983) | 5.887 1.416 24.465 3.30 Borison (1992) | 7.000 0.400 122.442 0.82 Chouinard (1993) | 3.492 1.113 10.955 5.13 Durost (1964) | 8.684 1.258 59.947 1.80 Garry (1962) | 1.721 0.569 5.203 5.47 Howard (1974) | 2.039 0.670 6.208 5.41 Marder (1994) | 1.348 0.736 2.467 18.33 Nishikawa (1982) | 3.000 0.137 65.903 0.70 Nishikawa (1984) | 9.281 0.583 147.831 0.87 Reschke (1974) | 3.793 1.058 13.604 4.11 Selman (1976) | 1.682 0.639 4.432 7.14 Serafetinides (1972) | 7.322 0.406 131.937 0.80 Simpson (1967) | 2.042 0.102 40.743 0.75 Spencer (1992) | 11.000 1.671 72.396 1.89 Vichaiya (1971) | 17.035 0.996 291.362 0.83 ---------------------+--------------------------------------------------- I-V pooled ES | 1.875 1.447 2.429 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 20.45 (d.f. = 16) p = 0.201 I-squared (variation in ES attributable to heterogeneity) = 21.7% Test of ES=1 : z= 4.76 p = 0.000 . . . // Part 2: Mirtazapine meta-analysis . . * download metamiss2, if required . *net from http://missoptima.project.uoi.gr/ . *net install metamiss2 . . use mirtazapine, clear (Mirtazapine meta-analysis: outcome is change in depression symptoms) . . * Q12. Start by drawing a forest plot and doing a simple meta-analysis (ACA) using . metan nt yt sdt nc yc sdc, nostandard lcols(study) randomi Study | WMD [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- Claghorn 1995 | -3.100 -8.799 2.599 9.37 MIR 003-003 | -2.500 -6.814 1.814 12.37 MIR 003-008 | -1.200 -6.215 3.815 10.75 MIR 003-020 | -6.800 -11.305 -2.295 11.91 MIR 003-021 | 3.600 0.251 6.949 14.92 MIR 003-024 | -4.600 -9.038 -0.162 12.07 MIR 84023a | -2.300 -6.166 1.566 13.52 MIR 84023b | -2.900 -6.191 0.391 15.09 ---------------------+--------------------------------------------------- D+L pooled WMD | -2.305 -4.619 0.009 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 17.02 (d.f. = 7) p = 0.017 I-squared (variation in WMD attributable to heterogeneity) = 58.9% Estimate of between-study variance Tau-squared = 6.4208 Test of WMD=0 : z= 1.95 p = 0.051 . . * Q13. Use metamiss2 to replicate the results from the metan command (conduct an ACA): . metamiss2 nt mt yt sdt nc mc yc sdc, impmean(0 0) impsd(0 0) ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Available cases analysis ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- 1 | -3.100 -8.799 2.599 9.37 2 | -2.500 -6.814 1.814 12.37 3 | -1.200 -6.215 3.815 10.75 4 | -6.800 -11.305 -2.295 11.91 5 | 3.600 0.251 6.949 14.92 6 | -4.600 -9.038 -0.162 12.07 7 | -2.300 -6.166 1.566 13.52 8 | -2.900 -6.191 0.391 15.09 ---------------------+--------------------------------------------------- D+L pooled ES | -2.305 -4.619 0.009 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 17.02 (d.f. = 7) p = 0.017 I-squared (variation in ES attributable to heterogeneity) = 58.9% Estimate of between-study variance Tau-squared = 6.4208 Test of ES=0 : z= 1.95 p = 0.051 . . * Q14. MNAR centred at MAR. . metamiss2 nt mt yt sdt nc mc yc sdc, impmean(0 0) impsd(1.5 1.5) ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Bayesian analysis using priors ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Priors used: Experimental group: N(0,1.5^2); Control group: N(0,1.5^2); Correlation: 0 Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- 1 | -3.100 -9.182 2.982 8.85 2 | -2.500 -7.187 2.187 12.36 3 | -1.200 -6.692 4.292 10.16 4 | -6.800 -11.704 -1.896 11.72 5 | 3.600 -0.513 7.713 14.25 6 | -4.600 -9.395 0.195 12.04 7 | -2.300 -6.549 1.949 13.78 8 | -2.900 -6.346 0.546 16.83 ---------------------+--------------------------------------------------- D+L pooled ES | -2.348 -4.514 -0.181 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 12.36 (d.f. = 7) p = 0.089 I-squared (variation in ES attributable to heterogeneity) = 43.4% Estimate of between-study variance Tau-squared = 4.1664 Test of ES=0 : z= 2.12 p = 0.034 . . metamiss2 nt mt yt sdt nc mc yc sdc, compare(impmean(0 0) impsd(1.5 1.5)) /// > impmean(0 0) impsd(0 0) Primary analysis ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Available cases analysis ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Secondary analysis ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Bayesian analysis using priors ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Priors used: Experimental group: N(0,1.5^2); Control group: N(0,1.5^2); Correlation: 0 Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Study | ES [95% Conf. Interval] ---------------------+--------------------------------------------------- Primary analysis 1 | -3.100 -8.799 2.599 2 | -2.500 -6.814 1.814 3 | -1.200 -6.215 3.815 4 | -6.800 -11.305 -2.295 5 | 3.600 0.251 6.949 6 | -4.600 -9.038 -0.162 7 | -2.300 -6.166 1.566 8 | -2.900 -6.191 0.391 Sub-total | D+L pooled ES | -2.305 -4.619 0.009 ---------------------+--------------------------------------------------- Secondary analysis 9 | -3.100 -9.182 2.982 10 | -2.500 -7.187 2.187 11 | -1.200 -6.692 4.292 12 | -6.800 -11.704 -1.896 13 | 3.600 -0.513 7.713 14 | -4.600 -9.395 0.195 15 | -2.300 -6.549 1.949 16 | -2.900 -6.346 0.546 Sub-total | D+L pooled ES | -2.348 -4.514 -0.181 ---------------------+--------------------------------------------------- Test(s) of heterogeneity: Heterogeneity degrees of statistic freedom P I-squared** Tau-squared Primary analysis 17.02 7 0.017 58.9% 6.4208 Secondary analysis 12.36 7 0.089 43.4% 4.1664 ** I-squared: the variation in ES attributable to heterogeneity) Significance test(s) of ES=0 Primary analysis z= 1.95 p = 0.051 Secondary analysis z= 2.12 p = 0.034 ------------------------------------------------------------------------- . . * Q15. MNAR centred away from MAR. . metamiss2 nt mt yt sdt nc mc yc sdc, impmean(0.5 -1) impsd(1 1.5) ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Bayesian analysis using priors ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Priors used: Experimental group: N(0.5,1^2); Control group: N(-1,1.5^2); Correlation: 0 Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- 1 | -2.311 -8.323 3.701 9.07 2 | -1.833 -6.438 2.771 12.36 3 | -0.458 -5.783 4.867 10.53 4 | -6.120 -10.910 -1.329 11.86 5 | 4.460 0.534 8.386 14.37 6 | -3.940 -8.654 0.774 12.06 7 | -1.671 -5.820 2.479 13.68 8 | -2.526 -5.943 0.892 16.06 ---------------------+--------------------------------------------------- D+L pooled ES | -1.678 -3.937 0.580 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 14.04 (d.f. = 7) p = 0.051 I-squared (variation in ES attributable to heterogeneity) = 50.1% Estimate of between-study variance Tau-squared = 5.2258 Test of ES=0 : z= 1.46 p = 0.145 . . * Q16. Add correlation. . metamiss2 nt mt yt sdt nc mc yc sdc, impmean(0.5 1) impsd(1 1.5) impcorr(0.5) ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Bayesian analysis using priors ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Priors used: Experimental group: N(0.5,1^2); Control group: N(1,1.5^2); Correlation: 0.5 Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- 1 | -3.467 -9.360 2.427 9.05 2 | -2.767 -7.253 1.719 12.33 3 | -1.325 -6.503 3.853 10.58 4 | -7.003 -11.666 -2.341 11.86 5 | 3.300 -0.380 6.980 14.74 6 | -4.860 -9.460 -0.260 12.02 7 | -2.513 -6.539 1.513 13.66 8 | -3.071 -6.442 0.299 15.76 ---------------------+--------------------------------------------------- D+L pooled ES | -2.551 -4.781 -0.320 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 14.53 (d.f. = 7) p = 0.043 I-squared (variation in ES attributable to heterogeneity) = 51.8% Estimate of between-study variance Tau-squared = 5.2634 Test of ES=0 : z= 2.24 p = 0.025 . . * Q17. Increasing SD of impmean. . metamiss2 nt mt yt sdt nc mc yc sdc, impmean(0 0) impsd(2 2) ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Bayesian analysis using priors ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Priors used: Experimental group: N(0,2^2); Control group: N(0,2^2); Correlation: 0 Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- 1 | -3.100 -9.465 3.265 8.50 2 | -2.500 -7.458 2.458 12.31 3 | -1.200 -7.037 4.637 9.72 4 | -6.800 -11.993 -1.607 11.53 5 | 3.600 -1.021 8.221 13.54 6 | -4.600 -9.655 0.455 11.98 7 | -2.300 -6.824 2.224 13.92 8 | -2.900 -6.462 0.662 18.50 ---------------------+--------------------------------------------------- D+L pooled ES | -2.393 -4.484 -0.301 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 10.25 (d.f. = 7) p = 0.175 I-squared (variation in ES attributable to heterogeneity) = 31.7% Estimate of between-study variance Tau-squared = 2.8495 Test of ES=0 : z= 2.24 p = 0.025 . . metamiss2 nt mt yt sdt nc mc yc sdc, impmean(0 0) impsd(3 3) ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Bayesian analysis using priors ******** ******************************************************************* Model: IMDOM Effect measure: Mean difference Priors used: Experimental group: N(0,3^2); Control group: N(0,3^2); Correlation: 0 Method: Taylor series approximation Random effects meta-analysis (Calling metan ...) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- 1 | -3.100 -10.210 4.010 7.48 2 | -2.500 -8.161 3.161 11.80 3 | -1.200 -7.924 5.524 8.37 4 | -6.800 -12.742 -0.858 10.71 5 | 3.600 -2.233 9.433 11.12 6 | -4.600 -10.334 1.134 11.51 7 | -2.300 -7.531 2.931 13.82 8 | -2.900 -6.775 0.975 25.19 ---------------------+--------------------------------------------------- D+L pooled ES | -2.533 -4.478 -0.588 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 6.94 (d.f. = 7) p = 0.435 I-squared (variation in ES attributable to heterogeneity) = 0.0% Estimate of between-study variance Tau-squared = 0.0000 Test of ES=0 : z= 2.55 p = 0.011 . . * Q18. Full sensitivity analysis. . metamiss2 nt mt yt sdt nc mc yc sdc, sensitivity ******************************************************************* ******** METAMISS2: meta-analysis allowing for missing data ******* ******** Bayesian analysis using priors ******** **** Sensitivity analysis assuming departures from MAR ***** ******************************************************************* . end of do-file Log file practical.log completed on 3 Mar 2015 at 12:39:11 name: log: H:\missing\meta\Egger chapter\final\practical.log log type: text closed on: 3 Mar 2015, 12:39:11 -----------------------------------------------------------------------------------------------------------------------