Declaring data to be time-series data

[Filip Pierzgalski]

Hi everyone!

I have monthly data set presented as such

* Example generated by -dataex-. To install: ssc install dataex
clear
input long yyyymm double return
187101  .01242389
187102  .02019115
187103  .02964644
187104  .03405454
187105  .02848361
187106 -.01542547
187107 -.00863808
187108  .01915629
187109  .01136751
187110 -.05886838
187111  .03742423
187112   .0257288
187201  .02951672
187202  .00101633
187203  .04072984
187204  .03142682
187205 -.00213872
187206 -.00795212
187207 -.00034236
187208 -.00761387
187209 -.01335902
187210  .01832136
187211 -.00212152
187212  .03346823
187301  .01082923
187302  .01575765
187303 -.01307993
187304 -.00878622
187305  .01496919
187306 -.01535528
end
format %tm yyyymm

First i tried to simply use

tsset yyyymm, monthly

time variable:  yyyymm, 1.9e+05 to 2.0e+05, but with gaps
                delta:  1 month

But as you can see i get weird numbers out

I also tried do this:

 . gen date = ym(floor( yyyymm/100 ), mod( yyyymm, 100 ))  . format date %tm  . tsset date, monthly         time variable:  date, 1871m2 to 1873m7                 delta:  1 month

But later when i try to do:

newey return L.return, lag(10)

I get this syntax date is not regularly spaced

Any idea on what i should do in this situation? Thank you all in advance

[Nick Cox]

Your first approach was clearly wrong, as you don't have monthly data that are 187101 or more months from 1 January 1960, i.e. roughly 15,000 years into the future.

The second approach is clearly right, but I can't reproduce your problem with Stata 15.1.

. gen date = ym(floor( yyyymm/100 ), mod( yyyymm, 100 ))  

. format date %tm  

. tsset date, monthly        
        time variable:  date, 1871m1 to 1873m6
                delta:  1 month

.
. newey return L.return, lag(10)

Regression with Newey-West standard errors      Number of obs     =         29
maximum lag: 10                                 F(  1,        27) =       0.02
                                                Prob > F          =     0.8986

------------------------------------------------------------------------------
             |             Newey-West
      return |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      return |
         L1. |   .0146226   .1136528     0.13   0.899    -.2185736    .2478188
             |
       _cons |   .0084265   .0026522     3.18   0.004     .0029847    .0138683
------------------------------------------------------------------------------

That's a small sample to do much with, as presumably you realise.

(If you aren't using Stata 15.1, you should tell us that. https://www.statalist.org/forums/help#version applies.

[Joro Kolev]

There is something wrong here, and I do not see what it is.

You are correctly specifying your data as Stata monthly elapsed date. You might be even using a solution that I came up with at some point here ( https://www.statalist.org/forums/for...perfectly-fine ).

Then when you -tsset- your data, your command does not report any irregularities, it says delta is 1 month, it does not say irregularly spaced... But then -newey- claims that your data is irregularly spaced. This does not sound right. It sounds like this cannot be. Also this problem does not occur in the data that you apply with -dataex- above.

The only explanation that I can think of is that something has gone wrong with your data between the -tsset- command, and the -newey- command.

[Filip Pierzgalski]

Thank you for super fast answer, i'm using stata 15.1. I did exactly the same procedure, and i don't get the table

. gen date = ym(floor( yyyymm/100 ), mod( yyyymm, 100 ))

. format date %tm

. tsset date, monthly
        time variable:  date, 1871m1 to 2017m12
                delta:  1 month

. newey return L.return, lag(10)
date is not regularly spaced
r(198);

[Nick Cox]

#1 reports an error message that can't be reproduced for the example data only.

Now #4 reports an error message for data from 1871 to 2017 which we can't access.

You still need to show us a problem we can reproduce.

NOTE: Cross-posted at https://stackoverflow.com/questions/...eries-in-stata Our policy on cross-posting is explicit. You are asked to tell us about it.

[Filip Pierzgalski]

I also tried simply doing this:

. gen t=_n

. tsset t
        time variable:  t, 1 to 1764
                delta:  1 unit

. newey return L.return, lag(10)
t is not regularly spaced
r(198);

t is number from 1 to 1764, so how can this be not regulary spaced??

[Filip Pierzgalski]

I was missing 4 observation, it worked perfectly when i dropped them! Thank you for the help that i have recieved

[Joro Kolev]

Again, what you are doing is correct, and I often do this when I do not feel like fighting Stata dates, which I have found to be formidable foes.

. sort yyyymm

. gen mydate = _n

. tsset mydate
        time variable:  mydate, 1 to 30
                delta:  1 unit

. newey return l.return, lag(10)

Regression with Newey-West standard errors      Number of obs     =         29
maximum lag: 10                                 F(  1,        27) =       0.02
                                                Prob > F          =     0.8986

------------------------------------------------------------------------------
             |             Newey-West
      return |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      return |
         L1. |   .0146226   .1136528     0.13   0.899    -.2185736    .2478188
             |
       _cons |   .0084265   .0026522     3.18   0.004     .0029847    .0138683
------------------------------------------------------------------------------

And again this problem does not arise in the data you attached with -dataex-, and again I cannot explain from where the problem comes. Such a problem should not occur.

Can your -newey- ado be messed up for some reason? Why dont you try some other time series command that needs tsset to see whether the problem is not with -newey-?

[Nick Cox]

We can't help further, I guess, unless you post a complete reproducible example with the entire dataset that yields it. Use the options of dataex to override the default sample size.

EDIT: A revision to #7 appears to close the thread.