odkmeta writes a do-file to import ODK data to Stata, using the metadata from the survey and choices worksheets of the XLSForm. The do-file completes the following tasks in order:
- Import lists as value labels
- Add
othervalues to value labels - Import field attributes as characteristics
- Split
select_multiplevariables - Drop
notevariables - Format
date,time, anddatetimevariables - Attach value labels
- Attach field labels as variable labels and notes
- Merge repeat groups
The latest stable version is available through SSC: type ssc install odkmeta in Stata to install.
The certification script of odkmeta is cscript.do. If you are new to certification scripts, you may find this Stata Journal article helpful. See this guide for more on odkmeta testing.
Help keep odkmeta up-to-date by contributing code. See the issues for current to-dos. If you have experience with an ODK flavor not yet supported, please contribute your knowledge through pull requests or issues.
The Mata class `ODKMetaController' writes the odkmeta do-file, and odkmeta is largely just a wrapper of `ODKMetaController'. odkmeta uses the Stata Template Engine, which means that `ODKMetaController' writes from the do-file section templates in templates/. Most contributors will wish to modify one of the templates or add a new one.
When contributing code, adding associated cscript tests is much appreciated.
Follow [these steps](/doc/develop/Development environment.md) to set up your Stata environment for odkmeta development.
See LICENSE.txt for licensing information (MIT).
Converted automatically from SMCL:
log html odkmeta.sthlp odkmeta.md
The help file looks best when viewed in Stata as SMCL.
Titleodkmeta -- Create a do-file to import ODK data
odkmeta using filename, csv(csvfile) survey(surveyfile, surveyopts) choices(choicesfile, choicesopts) [options]
options Description ------------------------------------------------------------------------- Main * csv(csvfile) name of the .csv file that contains the ODK data * survey(surveyfile, surveyopts) import metadata from the survey worksheet surveyfile * choices(choicesfile, choicesopts) import metadata from the choices worksheet choicesfile
Fields dropattrib(headers) do not import field attributes with the column headers headers keepattrib(headers) import only field attributes with the column headers headers relax ignore fields in surveyfile that do not exist in csvfile
Lists other(other) Stata value of other values of select or_other fields; default is max oneline write each list on a single line
Options replace overwrite existing filename ------------------------------------------------------------------------- * csv(), survey(), and choices() are required.
surveyopts Description ------------------------------------------------------------------------- Main type(header) column header of the type field attribute; default is type name(header) column header of the name field attribute; default is name label(header) column header of the label field attribute; default is label disabled(header) column header of the disabled field attribute; default is disabled -------------------------------------------------------------------------
choicesopts Description ------------------------------------------------------------------------- Main listname(header) column header of the list_name list attribute; default is list_name name(header) column header of the name list attribute; default is name label(header) column header of the label list attribute; default is label -------------------------------------------------------------------------
other Description ------------------------------------------------------------------------- max maximum value of each list: maximum list value plus one min minimum value of each list: minimum list value minus one # constant value for all value labels -------------------------------------------------------------------------
odkmeta creates a do-file to import ODK data, using the metadata from the survey and choices worksheets of the XLSForm. The do-file, saved to filename, completes the following tasks in order:
o Import lists as value labels o Add other values to value labels o Import field attributes as characteristics o Split select_multiple variables o Drop note variables o Format date, time, and datetime variables o Attach value labels o Attach field labels as variable labels and notes o Merge repeat groups
After select_multiple variables have been split, tasks can be removed from the do-file without affecting other tasks. User-written supplements to the do-file may make use of any field attributes, which are imported as characteristics.
The odkmeta do-file uses insheet to import data. Fields that are long strings of digits, such as simserial fields, will be imported as numeric even if they are more than 16 digits. As a result, they will lose precision.
The do-file makes limited use of Mata to manage variable labels, value labels, and characteristics and to import field attributes and lists that contain difficult characters.
The do-file starts with the definitions of several local macros; these are constants that the do-file uses. For instance, local macro `datemask' is the mask of date values in the .csv files. The local macros are automatically set to default values, but they may need to be changed depending on the data.
ODK field names follow different conventions from Stata's constraints on variable names. Further, the field names in the .csv files are the fields' "long names," which are formed by concatenating the list of the groups in which the field is nested with the field's "short name." ODK long names are often much longer than the length limit on variable names, which is 32 characters.
These differences in convention lead to three kinds of problematic field names:
1. Long field names that involve an invalid combination of characters, for example, a name that begins with a colon followed by a number. insheet will not convert these to Stata names, instead naming each variable v concatenated with a positive integer, for example, v1. 2. Long field names that are unique ODK names but when converted to Stata names and truncated to 32 characters become duplicates. insheet will again convert these to v# names. 3. Long field names of the form v# that become duplicates with other variables that cannot be converted, for which insheet chooses v# names. These will be converted to different v# names.
Because of problem 3, it is recommended that you do not name fields as v#.
If a field name cannot be imported, its characteristic Odk_bad_name is 1; otherwise it is 0.
Most tasks that the odkmeta do-file completes do not depend on variable names. There are two exceptions:
1. The do-file uses split to split select_multiple variables. split will result in an error if a select_multiple variable has a long name or if splitting it would result in duplicate variable names. 2. The do-file uses reshape and merge to merge repeat groups. reshape will result in an error if there are long variable names. The merging code will result in an error if there are duplicate variable names in two datasets.
Where variable names result in an error, renaming is left to the user. The section of the do-file for splitting is preceded by a designated area for renaming. In the section for reshaping and merging, each repeat group has its own area for renaming.
Many forms do not require any variable renaming. For others, only a few variables need to be renamed; such renaming should go in the designated areas. However, some forms, usually because of many nested groups or groups with long names, have many long field names that become duplicate Stata names (problem 2 above). In this case, it may work best to use fields' short names where possible. The following code attempts to rename variables to their field short names. Place it as-is before the renaming for split:
foreach var of varlist _all { if "`:char `var'[Odk_group]'" != "" { local name = "`:char `var'[Odk_name]'" + /// cond(`:char `var'[Odk_is_other]', "_other", "") + /// "`:char `var'[Odk_geopoint]'" local newvar = strtoname("`name'") capture rename `var' `newvar' } }
ODK list names are not necessarily valid Stata names. However, odkmeta uses list names as value label names, and it requires that all ODK list names be Stata names.
ODK lists are lists of associations of names and labels. There are two broad categories of lists: those whose names are all integer and those with at least one noninteger name. In the former case, the values of the value label are the same as the names of the list. In the latter, the values of the value label indicate the order of the names within the list: the first name will equal 1, the second 2, and so on. For such lists, the value of the value label may differ from the name of the list even if the name is a valid value label value; what matters is whether all names of the list are integer.
However, the value labels of these lists are easy to modify. Simply change the values of the value labels in the do-file; the rest of the do-file will be unaffected. Do not change the value label text.
Certain names do not interact well with insheet, which the odkmeta do-file uses to import the data.
For instance, it is not always possible to distinguish a name of "." from sysmiss. When it is unclear, the do-file assumes that values equal the name "." and not sysmiss. The problem arises when insheet imports select fields whose names in the data are the same as the values of a Stata numeric variable: real numbers, sysmiss, and extended missing values. insheet imports such fields as numeric, converting blank values ("") as sysmiss, thereby using the same Stata value for the name "." and for blank values.
insheet does not always interact well with list values' names that look like numbers with leading zeros, for example, 01 or 0.5. If insheet imports a select field as numeric, it will remove such leading zeros, leading to incorrect values or an error in the do-file. For similar reasons, trailing zeros after a decimal point may be problematic.
List values' names that look like decimals may also not interact well with insheet. If insheet imports a select field as numeric, the do-file will convert it to string. However, for precision reasons, the resulting string may differ from the original name if the decimal has no exact finite-digit representation in binary.
Generally, names that look like numbers that cannot be stored precisely as double are problematic. This includes numbers large in magnitude.
odkmeta is not designed for features specific to ODK variants, such as SurveyCTO or formhub. However, it is often possible to modify the odkmeta do-file to account for these features, especially as all field attributes are imported as characteristics.
SurveyCTO
For instance, the odkmeta do-file will result in an error for SurveyCTO forms that contain dynamic choice lists. One solution is to make the following changes to the do-file in order to import select fields with dynamic lists as string variables.
One section of the odkmeta do-file encodes select fields whose list contains a noninteger name. Here, remove dynamic lists from the list of such lists:
* Encode fields whose list contains a noninteger name. local lists list1 list2 list3 ... ...
Above, if list3 were a dynamic list, it should be removed.
The next section of the do-file attaches value labels to variables:
* Attach value labels. ds, not(vallab) if "`r(varlist)'" != "" /// ds `r(varlist)', has(char Odk_list_name) foreach var in `r(varlist)' { if !`:char `var'[Odk_is_other]' { ...
Add a line to the second if command to exclude fields whose appearance attribute contains a search() expression:
* Attach value labels. ds, not(vallab) if "`r(varlist)'" != "" /// ds `r(varlist)', has(char Odk_list_name) foreach var in `r(varlist)' { if !`:char `var'[Odk_is_other]' & /// !strmatch("`:char `var'[Odk_appearance]'", "*search(*)*") { ...
The do-file will now import fields with dynamic lists without resulting in an error.
formhub
formhub does not export note fields in the .csv files; specify option relax to odkmeta.
formhub exports blank values as "n/a". Multiple sections of the odkmeta do-file must be modified to accommodate these.
Immediately before this line in the section for formatting date, time, and datetime variables:
if inlist("`type'", "date", "today") {
add the following line:
replace `var' = "" if `var' == "n/a"
Immediately before this line in the section for attaching value labels:
replace `var' = ".o" if `var' == "other"
add the following line:
replace `var' = "" if `var' == "n/a"
These lines replace "n/a" values with blank ("").
Remarks for "don't know," refusal, and other missing values
ODK lists may contain missing values, including "don't know" and refusal values. These will be imported as nonmissing in Stata. However, if the lists use largely consistent names or labels for the values, it may be possible to automate the conversion of the values to extended missing values in Stata. The following SSC programs may be helpful:
labmvs ssc install labutil2 labmv ssc install labutil2 labrecode ssc install labutil2 labelmiss ssc install labelmiss
+------+ ----+ Main +-------------------------------------------------------------
survey(surveyfile, surveyopts) imports the field metadata from the XLSForm's survey worksheet. survey() requires surveyfile to be a comma-separated text file. Strings with embedded commas, double quotes, or end-of-line characters must be enclosed in quotes, and embedded double quotes must be preceded by another double quote.
Each attribute in the survey worksheet has its own column and column header. Use the suboptions type(), name(), label(), and disabled() to specify alternative column headers for the type, name, label, and disabled attributes, respectively. All field attributes are imported as characteristics.
If the survey worksheet has duplicate column headers, only the first column for each column header is used.
The type characteristic is standardized as follows:
o select one is replaced as select_one. o select or other is replaced as select or_other: select_one list_name or other is replaced as select_one list_name or_other, and select_multiple list_name or other is replaced as select_multiple list_name or_other. o begin_group is replaced as begin group; end_group is replaced as end group; begin_repeat is replaced as begin repeat; and end_repeat is replaced as end repeat.
In addition to the attributes specified in the survey worksheet, odkmeta attaches these characteristics to variables:
Odk_bad_name is 1 if the variable's name differs from its ODK field name and 0 if not. See the remarks for field names above.
Odk_group contains a list of the groups in which the variable is nested, in order of the group level.
Odk_long_name contains the field's "long name," which is formed by concatenating the list of the groups in which the field is nested with the field "short name," with elements separated by "-".
Odk_repeat contains the (long) name of the repeat group in which the variable is nested.
Odk_list_name contains the name of a select field's list.
Odk_or_other is 1 if the variable is a select or_other field and 0 if not.
Odk_is_other is 1 if the variable is a free-text other variable associated with a select or_other field; otherwise it is 0.
For geopoint variables, Odk_geopoint is the variable's geopoint component: Latitude, Longitude, Altitude, or Accuracy. For variables that are not type geopoint, Odk_geopoint is blank.
choices(choicesfile, choicesopts) imports the list metadata from the XLSForm's choices worksheet. choices() requires choicesfile to be a comma-separated text file. Strings with embedded commas, double quotes, or end-of-line characters must be enclosed in quotes, and embedded double quotes must be preceded by another double quote.
Each attribute in the choices worksheet has its own column and column header. Use the suboptions listname(), name(), and label() to specify alternative column headers for the list_name, name, and label attributes, respectively. List attributes are imported as value labels.
If the choices worksheet has duplicate column headers, only the first column for each column header is used.
+--------+ ----+ Fields +-----------------------------------------------------------
dropattrib(headers) specifies the column headers of field attributes that should not be imported as characteristics. _all specifies that all characteristics be dropped.
keepattrib(headers) specifies the column headers of field attributes to import as characteristics. _all means all column headers. Other attributes are not imported.
relax specifies that fields mentioned in surveyfile that do not exist in csvfile be ignored. By default, the do-file attempts to attach the characteristics to these variables, resulting in an error if the variable does not exist. For fields associated with multiple variables, for example, geopoint fields, relax attempts to attach the characteristics to as many variables as possible: an error does not result if some but not all variables exist.
+-------+ ----+ Lists +------------------------------------------------------------
other(other) specifies the Stata value of other values of select or_other fields.
max, the default, specifies that the Stata value of other vary by the field's list. For each list, other will be the maximum value of the list plus one.
min specifies that the Stata value of other vary by the field's list. For each list, other will be the minimum value of the list minus one.
# specifies a constant value for other that will be used for all lists.
oneline specifies that each list's value label definition be written on one line, rather than on multiple using #delimit ;.
+-------+ ----+ Other +------------------------------------------------------------
replace specifies that the odkmeta do-file be replaced if it already exists.
Create a do-file named import.do that imports ODK data, including the metadata in survey.csv and choices.csv . odkmeta using import.do, csv("ODKexample.csv") survey("survey.csv") choices("choices.csv")
Same as the previous odkmeta command, but specifies that the field name attribute appears in the fieldname column of survey_fieldname.csv . odkmeta using import.do, csv("ODKexample.csv") survey("survey_fieldname.csv", name(fieldname)) choices("choices.csv") replace
Same as the previous odkmeta command, but specifies that the list name attribute appears in the valuename column of choices_valuename.csv . odkmeta using import.do, csv("ODKexample.csv") survey("survey_fieldname.csv", name(fieldname)) choices("choices_valuename.csv", name(valuename)) replace
Create a do-file that imports all field attributes except for hint . odkmeta using import.do, csv("ODKexample.csv") survey("survey.csv") choices("choices.csv") dropattrib(hint) replace
Same as the previous odkmeta command, but does not import any field attributes . odkmeta using import.do, csv("ODKexample.csv") survey("survey.csv") choices("choices.csv") dropattrib(_all) replace
Create a do-file that imports other values of select or_other fields as 99 . odkmeta using import.do, csv("ODKexample.csv") survey("survey.csv") choices("choices.csv") other(99) replace
Lindsey Shaughnessy of Innovations for Poverty Action assisted in almost all aspects of odkmeta's development. She collaborated on the structure of the program, was a very helpful tester, and contributed information about ODK.
Matthew White
For questions or suggestions, submit a GitHub issue or e-mail [email protected].