<=, <, >, >=, !=, missing(), inrange(), inlist()

So far, we have dealt with the equality condition ==. The expression for “less than or equal to” is <=, “strictly less than” is <, “strictly more than” is >, and “more than or equal to” is >=. != stands for “not equal to”.

One short example: Instead of looking at the whole month of March, lets take a look at average daily covishield vaccinations at district level (all India sample) for:

• the first 15 days of March

summarize total_covishield if date_month == 3 & date_day <= 15

• between 16 and 25 March

summarize total_covishield if date_month == 3 & (date_day >= 16 & date_day <= 25)
summarize total_covishield if date_month == 3 & inrange(date_day, 16, 25)

• 26 through the end of March

summarize total_covishield if date_month == 3 & (date_day >= 26 & date_day <= 31)
summarize total_covishield if date_month == 3 & inrange(date_day, 26, 31)
summarize total_covishield if date_month == 3 & date_day >= 26
summarize total_covishield if date_month == 3 & date_day >= 26 & !missing(date_day)

Lots of heavy stuff happening in #2 and #3 above. Lets unpack:

• inrange(variable, lower value, upper value) is an alternative to writing variable <= lower value & variable >= upper value
• Not used in the examples but useful to know is the inlist() command. It can be used in place of multiple “or” conditions for a variable. Say, we want to look at the states of Punjab, Haryana, and Kerala. How would the conditional statement look? (state == "punjab" | state == "haryana" | state == "kerala") is lengthy and looks uncool. Can save time and space by using inlist(state,"punjab","haryana","kerala").
• In #3, since we know we need to go from March 26 to March 31 and there are no days beyond 31 (always a good idea to verify this in the data!), we can skip the date_day <= 31 bit. This is what I do in the third line of the code.
• Super important: Missing values in Stata are essentially infinity, so higher in value than non-missing values. What does this mean? If there are any missing values in the date_day variable, writing just date_day >= 26 will include them -> which we do not want! So, if you don’t want to define the upper limit but want to exclude missing observations of the date_day variable, always include the !missing(date_day) bit in your code -> which is what we do in the fourth line.

Sometimes, we want to condition on the missing values of a certain variable because this can give us valuable information about observations that do not have any value for that variable. In this case, our statement would look like if missing(variable). An often used shorthand for missing() is mi().

There is another way to identify missing values in Stata. Missing values for string variables are empty strings, so, for a missing state name, mi(state) and state == "" are equivalent. Missing values for numeric variables are much more complicated but you will typically not be dealing with the more complicated cases. Mostly, missing values for numeric values are represented by a dot .. So, for a missing value of age, mi(age) is roughly equivalent to age == .. Note that in 99.99% cases (I don’t have data to back this claim though), these are equivalent! They are not equivalent when the data contains extended missing values which are a way of differenting between reasons for missing values. These are .a, .b, … , .z. For example, a survey may code a missing value in income as .d if the respondent does not know the answer and as .r if the respondent refuses to answer. In this case, mi(income) identifies both the cases whereas income == . will identify none. You could alternatively write income == .d or income == .r as per what you desire to condition. For more details on extended missing values, refer here.