to.wide.RdFunction to convert data given in long format to a wide format.
a data frame in long format.
either the name (given as a character string) or the position (given as a single number) of the study variable in the data frame.
either the name (given as a character string) or the position (given as a single number) of the group variable in the data frame.
optional character string to specify the reference group (must be one of the groups in the group variable). If not given, the most frequently occurring group is used as the reference group.
either the names (given as a character vector) or the positions (given as a numeric vector) of the group-level variables.
a character string of length 2 giving the affix that is placed after the names of the group-level variables for the first and second group.
logical to specify whether a row id variable should be added to the data frame (the default is TRUE).
logical to specify whether a comparison id variable should be added to the data frame (the default is TRUE).
logical to specify whether a design id variable should be added to the data frame (the default is TRUE).
integer to specify the minimum length of the shortened group names for the comparison and design id variables (the default is 2).
character vector with three elements to specify the name of the id, comparison, and design variables (the defaults are "id", "comp", and "design", respectively).
A meta-analytic dataset may be structured in a ‘long’ format, where each row in the dataset corresponds to a particular study group (e.g., treatment arm). Using this function, such a dataset can be restructured into a ‘wide’ format, where each group within a study is contrasted against a particular reference group.
The study and group arguments are used to specify the study and group variables in the dataset (either as character strings or as numbers indicating the column positions of these variables in the dataset). Optional argument ref is used to specify the reference group (this must be one of the groups in the group variable). Argument grpvars is used to specify (either as a character vector or by giving the column positions) of those variables in the dataset that correspond to group-level outcomes (the remaining variables are treated as study-level outcomes).
The dataset is restructured so that a two-group study will yield a single row in the restructured dataset, contrasting the first group against the second/reference group. For studies with more than two groups (often called ‘multiarm’ or ‘multitreatment’ studies in the medical literature), the reference group is repeated as many times as needed (so a three-group study would yield two rows in the restructured dataset, contrasting two groups against a common reference group).
If a study does not include the reference group, then another group from the study will be used as the reference group. This group is chosen based on the factor levels of the grp variable (i.e., the last level that occurs in the study becomes the reference group).
To distinguish the names of the group-level outcome variables for the two first and second group in the restructured dataset, the strings given for the postfix argument are placed after the respective variable names.
If requested, row id, comparison id, and design id variables are added to the restructured dataset. The row id is simply a unique number for each row in the dataset. The comparison id variable indicates which two groups have been compared against each other). The design id variable indicates which groups were included in a particular study. The group names are shortened for the comparison and design variables (to at least minlen; the actual length might be longer to ensure uniqueness of the group names).
The examples below illustrate the use of this function.
A data frame with rows contrasting groups against a reference group and an appropriate number of columns (depending on the number of group-level outcome variables).
Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. Journal of Statistical Software, 36(3), 1–48. https://doi.org/10.18637/jss.v036.i03
contrmat for a function to construct a contrast matrix based on a dataset in wide format.
dat.hasselblad1998, dat.lopez2019, dat.obrien2003, dat.pagliaro1992, dat.senn2013 for illustrative examples.
### data in long format
dat <- dat.senn2013
dat <- dat[c(1,4,3,2,5,6)]
dat
#> study comment treatment ni mi sdi
#> 1 De Fronzo (1995) change metformin 213 -1.70 1.459
#> 2 De Fronzo (1995) change placebo 209 0.20 1.446
#> 3 Lewin (2007) change metformin 431 -0.74 1.106
#> 4 Lewin (2007) change placebo 144 0.08 1.004
#> 5 Willms (1999) change metformin 29 -2.50 0.862
#> 6 Willms (1999) change acarbose 31 -2.30 1.782
#> 7 Willms (1999) change placebo 29 -1.30 1.831
#> 8 Davidson (2007) change rosiglitazone 117 -1.20 1.097
#> 9 Davidson (2007) change placebo 116 0.14 1.093
#> 10 Wolffenbuttel (1999) change rosiglitazone 183 -0.90 1.100
#> 11 Wolffenbuttel (1999) change placebo 192 0.20 1.110
#> 12 Kipnes (2001) change pioglitazone 182 -1.20 1.369
#> 13 Kipnes (2001) change placebo 181 0.10 1.024
#> 14 Kerenyi (2004) change rosiglitazone 160 -0.91 0.990
#> 15 Kerenyi (2004) change placebo 154 -0.14 0.920
#> 16 Hanefeld (2004) raw pioglitazone 319 7.61 1.072
#> 17 Hanefeld (2004) raw metformin 320 7.45 1.073
#> 18 Derosa (2004) raw pioglitazone 45 6.80 0.800
#> 19 Derosa (2004) raw rosiglitazone 42 6.70 0.900
#> 20 Baksi (2004) change rosiglitazone 218 -1.20 1.112
#> 21 Baksi (2004) change placebo 233 0.10 1.036
#> 22 Rosenstock (2008) change rosiglitazone 59 -1.17 1.229
#> 23 Rosenstock (2008) change placebo 57 -0.08 1.208
#> 24 Zhu (2003) change rosiglitazone 210 -1.90 1.470
#> 25 Zhu (2003) change placebo 105 -0.40 1.300
#> 26 Yang (2003) change rosiglitazone 102 -1.09 1.650
#> 27 Yang (2003) change metformin 96 -0.95 1.500
#> 28 Vongthavaravat (2002) change rosiglitazone 164 -1.10 1.559
#> 29 Vongthavaravat (2002) change sulfonylurea 170 0.10 0.992
#> 30 Oyama (2008) change acarbose 41 -0.70 0.800
#> 31 Oyama (2008) change sulfonylurea 43 -0.30 0.600
#> 32 Costa (1997) change acarbose 36 -1.10 0.360
#> 33 Costa (1997) change placebo 29 -0.30 0.700
#> 34 Hermansen (2007) change sitagliptin 106 -0.30 0.940
#> 35 Hermansen (2007) change placebo 106 0.27 0.940
#> 36 Garber (2008) change vildagliptin 132 -0.63 1.034
#> 37 Garber (2008) change placebo 144 0.07 1.080
#> 38 Alex (1998) change metformin 291 0.13 1.428
#> 39 Alex (1998) change sulfonylurea 300 0.50 1.450
#> 40 Johnston (1994) change miglitol 68 -0.41 1.072
#> 41 Johnston (1994) change placebo 63 0.33 1.032
#> 42 Johnston (1998a) change miglitol 91 -0.43 0.954
#> 43 Johnston (1998a) change placebo 43 0.98 1.311
#> 44 Kim (2007) change rosiglitazone 57 -1.10 1.341
#> 45 Kim (2007) change metformin 56 -1.10 1.139
#> 46 Johnston (1998b) change miglitol 49 -0.12 1.400
#> 47 Johnston (1998b) change placebo 34 0.56 1.166
#> 48 Gonzalez-Ortiz (2004) change metformin 34 -1.30 1.861
#> 49 Gonzalez-Ortiz (2004) change placebo 37 -0.90 1.803
#> 50 Stucci (1996) raw benfluorex 28 8.03 1.290
#> 51 Stucci (1996) raw placebo 30 8.26 1.350
#> 52 Moulin (2006) change benfluorex 161 -0.82 1.028
#> 53 Moulin (2006) change placebo 156 0.19 1.374
### restructure to wide format
dat <- to.wide(dat, study="study", grp="treatment", ref="placebo", grpvars=4:6)
dat
#> study comment treatment.1 ni.1 mi.1 sdi.1 treatment.2 ni.2 mi.2 sdi.2 id
#> 1 Alex (1998) change metformin 291 0.13 1.428 sulfonylurea 300 0.50 1.450 1
#> 2 Baksi (2004) change rosiglitazone 218 -1.20 1.112 placebo 233 0.10 1.036 2
#> 3 Costa (1997) change acarbose 36 -1.10 0.360 placebo 29 -0.30 0.700 3
#> 4 Davidson (2007) change rosiglitazone 117 -1.20 1.097 placebo 116 0.14 1.093 4
#> 5 De Fronzo (1995) change metformin 213 -1.70 1.459 placebo 209 0.20 1.446 5
#> 6 Derosa (2004) raw pioglitazone 45 6.80 0.800 rosiglitazone 42 6.70 0.900 6
#> 7 Garber (2008) change vildagliptin 132 -0.63 1.034 placebo 144 0.07 1.080 7
#> 8 Gonzalez-Ortiz (2004) change metformin 34 -1.30 1.861 placebo 37 -0.90 1.803 8
#> 9 Hanefeld (2004) raw metformin 320 7.45 1.073 pioglitazone 319 7.61 1.072 9
#> 10 Hermansen (2007) change sitagliptin 106 -0.30 0.940 placebo 106 0.27 0.940 10
#> 11 Johnston (1994) change miglitol 68 -0.41 1.072 placebo 63 0.33 1.032 11
#> 12 Johnston (1998a) change miglitol 91 -0.43 0.954 placebo 43 0.98 1.311 12
#> 13 Johnston (1998b) change miglitol 49 -0.12 1.400 placebo 34 0.56 1.166 13
#> 14 Kerenyi (2004) change rosiglitazone 160 -0.91 0.990 placebo 154 -0.14 0.920 14
#> 15 Kim (2007) change metformin 56 -1.10 1.139 rosiglitazone 57 -1.10 1.341 15
#> 16 Kipnes (2001) change pioglitazone 182 -1.20 1.369 placebo 181 0.10 1.024 16
#> 17 Lewin (2007) change metformin 431 -0.74 1.106 placebo 144 0.08 1.004 17
#> 18 Moulin (2006) change benfluorex 161 -0.82 1.028 placebo 156 0.19 1.374 18
#> 19 Oyama (2008) change acarbose 41 -0.70 0.800 sulfonylurea 43 -0.30 0.600 19
#> 20 Rosenstock (2008) change rosiglitazone 59 -1.17 1.229 placebo 57 -0.08 1.208 20
#> 21 Stucci (1996) raw benfluorex 28 8.03 1.290 placebo 30 8.26 1.350 21
#> 22 Vongthavaravat (2002) change rosiglitazone 164 -1.10 1.559 sulfonylurea 170 0.10 0.992 22
#> 23 Willms (1999) change acarbose 31 -2.30 1.782 placebo 29 -1.30 1.831 23
#> 24 Willms (1999) change metformin 29 -2.50 0.862 placebo 29 -1.30 1.831 24
#> 25 Wolffenbuttel (1999) change rosiglitazone 183 -0.90 1.100 placebo 192 0.20 1.110 25
#> 26 Yang (2003) change metformin 96 -0.95 1.500 rosiglitazone 102 -1.09 1.650 26
#> 27 Zhu (2003) change rosiglitazone 210 -1.90 1.470 placebo 105 -0.40 1.300 27
#> comp design
#> 1 me-su me-su
#> 2 ro-pl ro-pl
#> 3 ac-pl ac-pl
#> 4 ro-pl ro-pl
#> 5 me-pl me-pl
#> 6 pi-ro pi-ro
#> 7 vi-pl vi-pl
#> 8 me-pl me-pl
#> 9 me-pi me-pi
#> 10 si-pl si-pl
#> 11 mi-pl mi-pl
#> 12 mi-pl mi-pl
#> 13 mi-pl mi-pl
#> 14 ro-pl ro-pl
#> 15 me-ro me-ro
#> 16 pi-pl pi-pl
#> 17 me-pl me-pl
#> 18 be-pl be-pl
#> 19 ac-su ac-su
#> 20 ro-pl ro-pl
#> 21 be-pl be-pl
#> 22 ro-su ro-su
#> 23 ac-pl ac-me-pl
#> 24 me-pl ac-me-pl
#> 25 ro-pl ro-pl
#> 26 me-ro me-ro
#> 27 ro-pl ro-pl
### data in long format
dat <- dat.hasselblad1998
dat
#> id study authors year trt xi ni
#> 1 1 1 Reid et al. 1974 no_contact 75 731
#> 2 2 1 Reid et al. 1974 ind_counseling 363 714
#> 3 3 2 Cottraux et al. 1983 no_contact 9 140
#> 4 4 2 Cottraux et al. 1983 ind_counseling 23 140
#> 5 5 2 Cottraux et al. 1983 grp_counseling 10 138
#> 6 6 3 Slama et al. 1990 no_contact 2 106
#> 7 7 3 Slama et al. 1990 ind_counseling 9 205
#> 8 8 4 Jamrozik et al. 1984 no_contact 58 549
#> 9 9 4 Jamrozik et al. 1984 ind_counseling 237 1561
#> 10 10 5 Rabkin et al. 1984 no_contact 0 33
#> 11 11 5 Rabkin et al. 1984 ind_counseling 9 48
#> 12 12 6 Decker and Evans 1989 self_help 20 49
#> 13 13 6 Decker and Evans 1989 ind_counseling 16 43
#> 14 14 7 Richmond et al. 1986 no_contact 3 100
#> 15 15 7 Richmond et al. 1986 ind_counseling 31 98
#> 16 16 8 Leung 1991 no_contact 1 31
#> 17 17 8 Leung 1991 ind_counseling 26 95
#> 18 18 9 Mothersill et al. 1988 self_help 11 78
#> 19 19 9 Mothersill et al. 1988 ind_counseling 12 85
#> 20 20 9 Mothersill et al. 1988 grp_counseling 29 170
#> 21 21 10 Langford et al. 1983 no_contact 6 39
#> 22 22 10 Langford et al. 1983 ind_counseling 17 77
#> 23 23 11 Gritz et al. 1992 no_contact 79 702
#> 24 24 11 Gritz et al. 1992 self_help 77 694
#> 25 25 12 Campbell et al. 1986 no_contact 18 671
#> 26 26 12 Campbell et al. 1986 self_help 21 535
#> 27 27 13 Sanders et al. 1989 no_contact 64 642
#> 28 28 13 Sanders et al. 1989 ind_counseling 107 761
#> 29 29 14 Hilleman et al. 1993 ind_counseling 12 76
#> 30 30 14 Hilleman et al. 1993 grp_counseling 20 74
#> 31 31 15 Gillams et al. 1984 ind_counseling 9 55
#> 32 32 15 Gillams et al. 1984 grp_counseling 3 26
#> 33 33 16 Mogielnicki et al. 1986 self_help 7 66
#> 34 34 16 Mogielnicki et al. 1986 grp_counseling 32 127
#> 35 35 17 Page et al. 1986 no_contact 5 62
#> 36 36 17 Page et al. 1986 ind_counseling 8 90
#> 37 37 18 Vetter and Ford 1990 no_contact 20 234
#> 38 38 18 Vetter and Ford 1990 ind_counseling 34 237
#> 39 39 19 Williams and Hall 1988 no_contact 0 20
#> 40 40 19 Williams and Hall 1988 grp_counseling 9 20
#> 41 41 20 Pallonen et al. 1994 no_contact 8 116
#> 42 42 20 Pallonen et al. 1994 self_help 19 149
#> 43 43 21 Russell et al. 1983 no_contact 95 1107
#> 44 44 21 Russell et al. 1983 ind_counseling 143 1031
#> 45 45 22 Stewart and Rosser 1982 no_contact 15 187
#> 46 46 22 Stewart and Rosser 1982 ind_counseling 36 504
#> 47 47 23 Russell et al. 1979 no_contact 78 584
#> 48 48 23 Russell et al. 1979 ind_counseling 73 675
#> 49 49 24 Kendrick et al. 1995 no_contact 69 1177
#> 50 50 24 Kendrick et al. 1995 ind_counseling 54 888
### restructure to wide format
dat <- to.wide(dat, study="study", grp="trt", ref="no_contact", grpvars=6:7)
dat
#> id study authors year trt.1 xi.1 ni.1 trt.2 xi.2 ni.2 comp design
#> 1 1 1 Reid et al. 1974 ind_counseling 363 714 no_contact 75 731 in-no in-no
#> 2 2 2 Cottraux et al. 1983 grp_counseling 10 138 no_contact 9 140 gr-no gr-in-no
#> 3 3 2 Cottraux et al. 1983 ind_counseling 23 140 no_contact 9 140 in-no gr-in-no
#> 4 4 3 Slama et al. 1990 ind_counseling 9 205 no_contact 2 106 in-no in-no
#> 5 5 4 Jamrozik et al. 1984 ind_counseling 237 1561 no_contact 58 549 in-no in-no
#> 6 6 5 Rabkin et al. 1984 ind_counseling 9 48 no_contact 0 33 in-no in-no
#> 7 7 6 Decker and Evans 1989 ind_counseling 16 43 self_help 20 49 in-se in-se
#> 8 8 7 Richmond et al. 1986 ind_counseling 31 98 no_contact 3 100 in-no in-no
#> 9 9 8 Leung 1991 ind_counseling 26 95 no_contact 1 31 in-no in-no
#> 10 10 9 Mothersill et al. 1988 grp_counseling 29 170 self_help 11 78 gr-se gr-in-se
#> 11 11 9 Mothersill et al. 1988 ind_counseling 12 85 self_help 11 78 in-se gr-in-se
#> 12 12 10 Langford et al. 1983 ind_counseling 17 77 no_contact 6 39 in-no in-no
#> 13 13 11 Gritz et al. 1992 self_help 77 694 no_contact 79 702 se-no se-no
#> 14 14 12 Campbell et al. 1986 self_help 21 535 no_contact 18 671 se-no se-no
#> 15 15 13 Sanders et al. 1989 ind_counseling 107 761 no_contact 64 642 in-no in-no
#> 16 16 14 Hilleman et al. 1993 grp_counseling 20 74 ind_counseling 12 76 gr-in gr-in
#> 17 17 15 Gillams et al. 1984 grp_counseling 3 26 ind_counseling 9 55 gr-in gr-in
#> 18 18 16 Mogielnicki et al. 1986 grp_counseling 32 127 self_help 7 66 gr-se gr-se
#> 19 19 17 Page et al. 1986 ind_counseling 8 90 no_contact 5 62 in-no in-no
#> 20 20 18 Vetter and Ford 1990 ind_counseling 34 237 no_contact 20 234 in-no in-no
#> 21 21 19 Williams and Hall 1988 grp_counseling 9 20 no_contact 0 20 gr-no gr-no
#> 22 22 20 Pallonen et al. 1994 self_help 19 149 no_contact 8 116 se-no se-no
#> 23 23 21 Russell et al. 1983 ind_counseling 143 1031 no_contact 95 1107 in-no in-no
#> 24 24 22 Stewart and Rosser 1982 ind_counseling 36 504 no_contact 15 187 in-no in-no
#> 25 25 23 Russell et al. 1979 ind_counseling 73 675 no_contact 78 584 in-no in-no
#> 26 26 24 Kendrick et al. 1995 ind_counseling 54 888 no_contact 69 1177 in-no in-no