dfuller - "no observations" error message

[Sean Fiedler]

I am attempting to run the dfuller command, but I get the error message "no observations". My data is tsset and has no missing observations. I did see a similar question posted on this forum but it did not have an answer. Thank you for your help.

freduse TB3MS if daten<td(31dec2009), clear
rename * , lower

tsset daten

tsline tb
dfuller tb3ms, lags(12)

[William Lisowski]

Your problem is that the user-written freduse command (available from SSC) creates daten as a daily date variable, when that FRED date is actually a monthly variable to which the first day of the month has been assigned.

. list in 1/5, clean

             date   tb3ms       daten  
  1.   1934-01-01     .72   01jan1934  
  2.   1934-02-01     .62   01feb1934  
  3.   1934-03-01     .24   01mar1934  
  4.   1934-04-01     .15   01apr1934  
  5.   1934-05-01     .16   01may1934

Note that your tsset command suggested this sort of problem with its warning about gaps and its indication that the delta was 1 day despite the apparently monthly nature of the start and end values.

. tsset daten
        time variable:  daten, 01jan1934 to 01dec2009, but with gaps
                delta:  1 day

You need to create a Stata monthly date variable from your daily date variable. The code below points the way.

. freduse TB3MS if daten<td(31dec2009), clear
(992 observations read)
(80 observations deleted)

. rename * , lower

.
. generate datem = mofd(daten)

. format datem %tm

. tsset datem
        time variable:  datem, 1934m1 to 2009m12
                delta:  1 month

. list in 1/5, clean

             date   tb3ms       daten    datem  
  1.   1934-01-01     .72   01jan1934   1934m1  
  2.   1934-02-01     .62   01feb1934   1934m2  
  3.   1934-03-01     .24   01mar1934   1934m3  
  4.   1934-04-01     .15   01apr1934   1934m4  
  5.   1934-05-01     .16   01may1934   1934m5  

.
. dfuller tb3ms, lags(12)

Augmented Dickey-Fuller test for unit root         Number of obs   =       899

                               ---------- Interpolated Dickey-Fuller ---------
                  Test         1% Critical       5% Critical      10% Critical
               Statistic           Value             Value             Value
------------------------------------------------------------------------------
 Z(t)             -1.995            -3.430            -2.860            -2.570
------------------------------------------------------------------------------
MacKinnon approximate p-value for Z(t) = 0.2886

.

[Sean Fiedler]

Works perfectly, thank you!