Collapse stata how to#

With reshape wide, the list of level one variables is a list of actual, existing variables, so you can use shortcuts like age-hispanic. then take what follows and store it in a new variable called person. The reshape long command will identify all the variables that start with age, race, etc. do not refer to individual variables, but to groups of variables, and j(person) does not refer to an existing variable at all. However, the meaning is quite different: age, race, etc. Reshape long age race maritalStatus edu income female hispanic, /// Reshaping from wide form to long form requires the exact same command, just replacing wide with long: The householdIncome variable (on the far right in the data browser) remains a single variable because it is a household level variable. Since the data set has to be rectangular, that means all the households have 16 of each level one variable, with most of them containing missing values. The largest household had 16 people in it, so storing the information for all of them required 16 of each variable. Note that there are 16 of each of the level one variables. The person variable has been changed from part of a compound row identifier to part of a compound column identifier. income) followed by the number of the person being described. Most of the variable names now have two parts: the name of the quantity described (e.g. As promised, there is now just one observation per household, and the household variable is now a unique identifier all by itself. Take a moment to look at the data browser. It calls them i and j rather than level one and level two because reshape can be used on data with more than two levels of hierarchy. The i() option specifies the level two identifier, which could be a compound identifier the j() option specifies the level one identifier. (Make sure this is true!) Finally, the command needs to know the identifiers. The variables you list are level one variables any variables you do not list are assumed to be level two variables. The reshape command always reshapes the entire data set, but to do so it needs to understand which variables are level one and which are level two. This is a bit different from the usual syntax where the list of variables tells the command which variables to act on. Then you give a list of all the level one variables. The syntax of the reshape command begins by specifying the form you want, in this case wide. Reshape wide age race maritalStatus edu income female hispanic, /// In wide form it would have one observation per household, and reshape can do that for you: The data set has one observation per person, or level one unit, so the data set is currently in long form. The variable householdIncome is a level two, or household-level variable, because any persons who live in the same household will have the same value of householdIncome. At this point we don't have any variables that describe the household, so let's make one before proceeding:īy household: egen householdIncome = total(income) The variables age, race, maritalStatus, edu, income, female, and hispanic all describe individual persons, making them level one variables. The level one and level two identifier variables are person and household, and you can confirm that they uniquely identify observations with: Recall that this is hierarchical data consisting of people living in households, so a person is a level one unit and a household is a level two unit. Start a do file that loads the cleaned version of the 2000 American Community Survey we worked with earlier:

In particular, you need to know the level one and level two identifiers, and which variables are level one variables and which are level two. The reshape command is very simple to use, if you understand the structure of the your data set.

Collapse stata how to#

We'll also learn how to turn a data set containing both level one and level two units into a data set containing just level two units.

We'll learn how reshape data so we can switch at will between the long form (one row per level one unit) and wide form (one row per level two unit). In this section we'll learn how to restructure data sets. This is part six of Data Wrangling in Stata.