Title
Makes retrievals from PSID real easy
Syntax psiduse || new_stub varname identifers [ || new_stub varname identifers ] using dirname [ , use_options ]
psidadd || new_stub varname identifers [ || new_stub varname identifers ] [ , add_options ]
dirname is the name of the directory in which the PSID files are stored. The term varname identifier refers to PSID variables. You cannot specify the PSID variable names in terms of a varlist but have to use the syntax specified below. Finally new_stub is the prefix of new names for variables that belong together.
options Description ------------------------------------------------------------------------- use_options design(designtype) Design; default: design(balanced) cnef(numlist) Waves to be used for CNEF data clear Replace data in memory correct Correct inconsistent 2005/2007 data delivery
add_options cneffrom(path) Path to CNEF data psidfrom(path) Path to PSID data correct Correct inconsistent 2005/2007 data delivery -------------------------------------------------------------------------
Description
psiduse and psidadd perform data retrievals from the Panel Study of Income Dynamics (PSID) and for the American part of the Cross National Equivalence File (CNEF). The programs are companions of soepuse and soepadd which provide a similar functionality for the German Socio Economic Panel.
The programs create PSID data sets holding the variables identified by the variable identifiers with names prefixed by new_stub. psiduse generates a new file and psidadd merges further variables to a file generated with psiduse. By default, the created files will have a balanced panel design, but various other designs could be specified.
To load data from the PSID, psiduse and psidadd require that the variables are specified very similar to the variable listing produced by the PSID Data Center. Here is an example: To create a longitudinal file with individual ages and subjective health evaluations of the household head of waves 1991 and 1992 you would specify
. psiduse || age [91]ER30692 [92]ER30736 || shealth [91]V20021 [92]V21321 using ~/data/psid05
or in a format that highlights better the requested format of the variable identifiers:
. psiduse || age [91]ER30692 [92]ER30736 || shealth [91]V20021 [92]V21321 using ~/data/psid05
This command will produce a longitudinal data set in a balanced panel design with variable names "age1991" and "age1992" for the age variables, and "shealth1991" and "shealth1992" for the health evaluations. The new data set will also contain person and housholds identifiers using the name conventions of the Cross National Equivalence File.
psiduse and psidadd are constructed for using them in connection with the PSID Data Center. Consider you have been using the PSID Data Center to search the PSID data base for items concerning health. After founding an item that suits your needs you have clicked on that item which brought up an item correspondence list that looks like this
[84]V10877 [85]V11991 [86]V13417 [87]V14513 [88]V15993 [89]V17390 [90]V18721 [91]V20021 [92]V21321 [93]V23180 [94]ER3853 [95]ER6723 [96]ER8969 [97]ER11723 [99]ER15447 [01]ER19612 [03]ER23009 [05]ER26990
This list almost completely resembles the format of the variable identifiers to be used in psiduse. It can be therefore copied into the command. Once you did this, you only need to add a name for the item. This name will be used as a prefix for all variable names created for that item in the new data set.
The entire psiduse command to load all variables of the example above will then become
. psiduse || health [84]V10877 [85]V11991 [86]V13417 [87]V14513 [88]V15993 [89]V17390 [90]V18721 [91]V20021 [92]V21321 [93]V23180 [94]ER3853 [95]ER6723 [96]ER8969 [97]ER11723 [99]ER15447 [01]ER19612 [03]ER23009 [05]ER26990 using ~/data/psid05
To load data from the CNEF, psiduse and psidadd require that the prefixes of variable names are listed as variable identifier, and that the option cnef() is specified. To load, for example, the pre- and post government incomes of waves 1980 to 1990 one would use
. psiduse || pre i11102 || post i11104 using ~/data/cnef, cnef(1980/1990)
Note that you cannot load CNEF variables and PSID variables with the same command. Either you use psiduse to load the CNEF variables and use psidadd to add variables from the PSID, or you do it the other way around.
Note also that you must not add variables from waves that are not already included in the file created by psiduse. If you use psidadd for adding CNEF data to an existing PSID data file, all waves that are included in the existing file are retained automatically.
Finnaly note that you must add a set of empty brackets in front of items that appear only once in the database (i.e. constants).
Options
design(designtype) specifies the design of the panel data to be created. design(balanced) is used to create a balanced panel design, i.e. the data will contain only observations interviewed in all requested waves. design(any) will keep all available observations in the data set. design(#) with # being an integer positive number creates data sets with households interviewed # times or more.
clear specifies that it is okay to replace the data in memory, even though the current data have not been saved to disk.
cnef(numlist) must be used to load data from the American part of the Cross National Equivalence File (CNEF). Specify the waves for which data should be retained inside the parentheses. The CNEF uses a standardized scheme for variable names which allows a simplified syntax for the specification of variable identifiers. The CNEF option lets you access this simplified syntax.
cneffrom(path) By default psidadd assumes that the data is stored in the directory specified by psiduse. If you want to add CNEF variables to a PSID data set you must specify the path to the CNEF data. You have to specify cneffrom() even if the CNEF data is stored in the PSID directory.
psidfrom(path) By default psidadd assumes that the data is stored in the directory specified by psiduse. If you want to add PSID variables to a CNEF data set you must specify the path to the PSID data. You have to specify psidfrom() even if the PSID data is stored in the CNEF directory.
correct An early version of the CNEF delivery for 2007 introduced upper cased "LL" in the variable names of three variables in the files for years 2005 and 2007. Moreover, the data file of 2005 contained 9 dublicate observations. Option correct changes "LL" to "ll" and removes the dublicates. I hope that this option becomes superfluos with updated data deliveries.
Example(s)
Constructing Longitudinal Family Records (PSID) . psiduse || health [84]V10877 [85]V11991 [86]V13417 using .
Constructing Longitudinal Records (CNEF 1984-2005) . psiduse || pregov i11101 || postgov i11102 using . , cnef(1980(1)1995 1997(2)2005)
Linking Family and Individual Data (PSID) . psiduse || health [84]V10877 || age [84]ER30432 using .
A more practical example for a longitudinal data set with several items (PSID and CNEF)
. psiduse || shealth [84]V10877 [85]V11991 [86]V13417 [87]V14513 [88]V15993 [89]V17390 [90]V18721 [91]V20021 [92]V21321 [93]V23180 [94]ER3853 [95]ER6723 [96]ER8969 [97]ER11723 [99]ER15447 [01]ER19612 [03]ER23009 [05]ER26990 || age [68]ER30004 [69]ER30023 [70]ER30046 [71]ER30070 [72]ER30094 [73]ER30120 [74]ER30141 [75]ER30163 [76]ER30191 [77]ER30220 [78]ER30249 [79]ER30286 [80]ER30316 [81]ER30346 [82]ER30376 [83]ER30402 [84]ER30432 [85]ER30466 [86]ER30501 [87]ER30538 [88]ER30573 [89]ER30609 [90]ER30645 [91]ER30692 [92]ER30736 [93]ER30809 [94]ER33104 [95]ER33204 [96]ER33304 [97]ER33404 [99]ER33504 [01]ER33604 [03]ER33704 [05]ER33804 || disable [72]V2718 [73]V3244 [74]V3666 [75]V4145 [76]V4625 [77]V5560 [78]V6102 [79]V6710 [80]V7343 [81]V7974 [82]V8616 [83]V9290 [84]V10879 [85]V11993 [86]V13427 [87]V14515 [88]V15994 [89]V17391 [90]V18722 [91]V20022 [92]V21322 [93]V23181 [94]ER3854 [95]ER6724 [96]ER8970 [97]ER11724 [99]ER15449 [01]ER19614 [03]ER23014 [05]ER26995 using . , clear design(10) . psidadd || pregov i11101 || postgov i11102, cnef(~/data/cnef)
Note
psiduse and psidadd are two little unambitious helper programs. A far more advanced Stata program for working with large panel data sets is PanelWhiz by John Haisken DeNew.
Author
Ulrich Kohler, WZB, kohler@wzb.eu
Also see
Online: soepuse (if installed), rgroup (if installed)