The Date dimension was a time dimension in both versions; in one I didn't specify the format, in the other I did. The first one had klunky plots--I think because only the year was being plotted, or each quarter in a year was stacked vertically--while the latter had smooth plots.
The basic point is that I think we should aim to make specifying the date format unnecessary, unless there's a need to clarify when a US MDY format is used and it can't be distinguished from DMY formatted dates. In that case the plot should look the same whether or not the date format is specified for the date dimension.
I'll reproduce the two versions and video the the process and send them to you, though it will take time--really slow internet here in London.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I think I've worked out what's going wrong. When you don't specify the date format, the program seems to treat the Quarter numbers as months. So what you get is data for January, February, March, April, and then the next January-April. Instead what you should have is data for January, April, July, October, then the next January, April, July, October.
Of course what different countries choose as months for quarters is probably also arbitrary! But in this case I think it would be best to use the American standard.
In programming terms, Ravel needs to recognise a Q in a date string as indicating quarters, and therefore interpret the 01-04 as Q1-Q4.
AFAICT, quarterly data is just plain non-standard, and has to be handled by a user specified format string, as we do. The only standard form is ISO data format, which we handle in greater generality with the "blank" format string.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Yes, but what happens if we have two time series then, one with something that fits the ISO standard and one Quarterly? Will they be treated consistently, or will one bugger up the other?
Pardon my mood! My expensive USB microphone has just failed one day before the bloody YouTube Pitch. I'm now going to go back in time and kill Murphy.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
On Mon, Jul 15, 2024 at 07:45:53AM -0000, Steve Keen wrote:
Yes, but what happens if we have two time series then, one with something that
fits the ISO standard and one Quarterly? Will they be treated consistently, or
will one bugger up the other?
They will be treated consistently. At import, you'll need the correct
format for each series (blank in the first, %Y-Q%Q in the
second). Internally, they are represented by ptime objects, and
interpolation means the two series can be mixed and matched. The
format first used to define a dimension will be stored as an output
format, although that can be edited in the dimensions dialog. Output
formats are only relevant for sheet objects and CSV export. Plots
determine their own format that depends on the range - multiyear
ranges will display just the year, less than 2 years, the numbers will
include the month, less than two months will include the date and so
on.
Dr Russell Standish Phone 0425 253119 (mobile)
Principal, High Performance Coders hpcoder@hpcoders.com.au http://www.hpcoders.com.au
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
RVL and data files
The first plot has the x-vector set as string. The second plot has it set as time. This is expected.
What is actually being requested here?
The Date dimension was a time dimension in both versions; in one I didn't specify the format, in the other I did. The first one had klunky plots--I think because only the year was being plotted, or each quarter in a year was stacked vertically--while the latter had smooth plots.
The basic point is that I think we should aim to make specifying the date format unnecessary, unless there's a need to clarify when a US MDY format is used and it can't be distinguished from DMY formatted dates. In that case the plot should look the same whether or not the date format is specified for the date dimension.
I'll reproduce the two versions and video the the process and send them to you, though it will take time--really slow internet here in London.
I think I've worked out what's going wrong. When you don't specify the date format, the program seems to treat the Quarter numbers as months. So what you get is data for January, February, March, April, and then the next January-April. Instead what you should have is data for January, April, July, October, then the next January, April, July, October.
Of course what different countries choose as months for quarters is probably also arbitrary! But in this case I think it would be best to use the American standard.
In programming terms, Ravel needs to recognise a Q in a date string as indicating quarters, and therefore interpret the 01-04 as Q1-Q4.
AFAICT, quarterly data is just plain non-standard, and has to be handled by a user specified format string, as we do. The only standard form is ISO data format, which we handle in greater generality with the "blank" format string.
Yes, but what happens if we have two time series then, one with something that fits the ISO standard and one Quarterly? Will they be treated consistently, or will one bugger up the other?
Pardon my mood! My expensive USB microphone has just failed one day before the bloody YouTube Pitch. I'm now going to go back in time and kill Murphy.
On Mon, Jul 15, 2024 at 07:45:53AM -0000, Steve Keen wrote:
They will be treated consistently. At import, you'll need the correct
format for each series (blank in the first, %Y-Q%Q in the
second). Internally, they are represented by ptime objects, and
interpolation means the two series can be mixed and matched. The
format first used to define a dimension will be stored as an output
format, although that can be edited in the dimensions dialog. Output
formats are only relevant for sheet objects and CSV export. Plots
determine their own format that depends on the range - multiyear
ranges will display just the year, less than 2 years, the numbers will
include the month, less than two months will include the date and so
on.
Dr Russell Standish Phone 0425 253119 (mobile)
Principal, High Performance Coders hpcoder@hpcoders.com.au
http://www.hpcoders.com.au