Milliseconds problem with clock

[Andreas Loeffler]

I am using Stata/SE 17.0 for Mac (Apple Silicon, Revision 05 Oct 2021) with a trivial example. My raw data (these are bid-ask spreads) look as follows

 
2021-11-23 17:45:00.154
64
UP
249,85
249,95
147

2021-11-23 17:35:16.634

UP

249,45
1.837

2021-11-23 17:35:16.634

UP

249,45
1.694

2021-11-23 17:35:16.633

UP

249,45
1.037

No, I use the usual way to convert the timestamp, for example I do

gen double timestamp=clock(var1, "YMDhms")
format timestamp %tc

and indeed I get

var1    bid    ask    spread    timestamp
2021-11-15 09:00:13.058    253.5    254.75    1.25    15nov2021 09:00:13
2021-11-15 09:00:13.140    253.75    253.8    .05    15nov2021 09:00:13
2021-11-15 09:00:13.140    253.8    254.65    .85    15nov2021 09:00:13

As can be seen the milliseconds from the string variable are probably not converted to Stata. In particular, the following shows this

 tsset timestamp
repeated time values in sample

Does anyone know what to do?

[Leonardo Guizzetti]

The variable -timestamp- does actually contain the milliseconds, so you have converted the time faithfully into Stata's internal format. To convince yourself, study this example.

. di %tc clock("2021-11-23 17:45:00.154", "YMDhms")
23nov2021 17:45:00

. di %tcDDmonCCYY_HH:MM:SS.sss clock("2021-11-23 17:45:00.154", "YMDhms")
23nov2021 17:45:00.154

As for the error you received:

repeated time values in sample

Stata has told you the problem, and now you know it is nothing to do with the time conversion. Stata told you that you have repeated time values, and by definition of a time series, you cannot have repeated times. If you have problems with a time-series analysis, this would probably be better for a new thread.

[Andreas Loeffler]

I did not know that the milliseconds are contained (but not shown). There were some duplicates with even the same milliseconds and after I removed them it now works. Unfortunately, the data seem to be of bad quality now...