Centered moving average in panel data

[Claire McKenna]

How do I create a centered moving average in panel data? I've consulted this document: https://www.stata.com/support/faqs/s...nd-panel-data/ -- specifically the section on panel data. But I'm not sure how / whether to use to tsset. I want to take the centered moving average of a variable over five consecutive months, across my main time variable: year_month. Thanks very much.

[Clyde Schechter]

First, your time variable must be a true Stata internal format (monthly) date variable. If it isn't, your first task is to create such a variable.

With that behind you, having panel data you will want to xtset your data. (In fact, if you try to -tsset- it you will probably get error messages as multiple observations with the same time will be detected.)

Once you have -xtset- your data, it's just

tssmooth ma wanted = variable_to_be_smoothed, window(-2 1 2)

[Nick Cox]

The FAQ cited in #1 is flagged as for users of Stata 7.

Essentially if you

xtset identifier time

or

tsset identifier time

then tssmooth will do the right thing.

Another way to do it is with rangestat (SSC). This doesn't require, or check for, declaring your data as time series or panel data, but it can happily be used with such data.

. webuse grunfeld, clear

. xtset

Panel variable: company (strongly balanced)
 Time variable: year, 1935 to 1954
         Delta: 1 year

. tssmooth ma mean=invest , w(2 1 2)
The smoother applied was
     by company : (1/5)*[x(t-2) + x(t-1) + 1*x(t) + x(t+1) + x(t+2)]; x(t)= invest

. rangestat (mean) mean_rs = invest, i(year -2 2) by(company)

. scatter mean_rs mean

. d

Contains data from https://www.stata-press.com/data/r17/grunfeld.dta
 Observations:           200                  
    Variables:             8                  3 Mar 2020 20:27
-----------------------------------------------------------------------------------------
Variable      Storage   Display    Value
    name         type    format    label      Variable label
-----------------------------------------------------------------------------------------
company         byte    %9.0g                 Company
year            int     %ty                   Year
invest          float   %9.0g                 Investment (current year)
mvalue          float   %9.0g                 Market value (prior year)
kstock          float   %9.0g                 Capital stock (prior year)
time            byte    %9.0g                 Time
mean            float   %9.0g                 ma: x(t)= invest : window(2 1 2)
mean_rs         double  %10.0g                mean of invest
-----------------------------------------------------------------------------------------
Sorted by: company  year
     Note: Dataset has changed since last saved.

Detail: rangestat produces doubles and tssmooth floats by default. For most applications, any differences will be immaterial.

[Claire McKenna]

Thank you both very much.