Another wishlist: focussing on some quirks and UI

[Tatha]

Hi Gareth,

Great to see that an Eviews forum has now become active. When I first began writing to you by e-mail on various things, a forum had not yet been started and no user-created library of programs was available. So, congratulations on launching it, and its great to see that someone knowledgeable like yourself is actively moderating.

Over those many private e-mails shared with you and colleagues regarding features and requests, I ended up putting together some sort of feature wish-list. Some of the requests have since been confirmed to be in Eviews 7, while others aren’t, though these may prove interesting for further subsequent versions. With this in mind, I thought it would be good to recap the wish-list here on the forum, so others can see and build on it. Wish-lists are essential in any software forum... :D

However, at the outset, let me state that my use of Eviews is slightly different from what I sense many others’ is – while I do some modeling alright, this tends to be in the form of one-off exercises/projects which add to an otherwise growing body of maintenance work (quick updates and commentary). During the modeling exercises, I can live with features which are not perfectly optimized, as I’m willing to manually intervene more. My focus, though, is more ‘applied’, like you might imagine goes on at policy bodies, central banks and practicing high-frequency desks (i.e. quick graph and comment on latest industrial production data etc etc.). And the main emphasis in my feature requests is ‘automation’.

I’ll start with the ones which are more housekeeping than full feature; and the suggested modification would possibly benefit all users (i.e. I guess that the current behaviour is probably not a specific design choice). These are bullets 1-6. I make a few feature suggestions later (bullets 7-12). Of course, if a later suggestion is adopted, a number of smaller ones could get subsumed by it.

And, of course, you will undoubtedly tell me that most of these are already possible; I’m just trying it the wrong way!....:) If so, I need to know.

1. Graph lines/area: tick-marks (inside axis) and axis zero-lines are currently hidden by lines and areas, if they overlap. Usually, one would want these to show on top of other graphics. This may already be corrected in Eviews 7 as per your colleague Pamela’s e-mail reply to me.

2. Graph overwrite: when we name an object the same as an existing workfile object, the program quits with an error. When we copy an object with the same name, however, we get the option to ‘overwrite’/rename etc. Would be good if the option to overwrite was offered even while naming – by dialog box in interactive mode and by an ‘o’ option in a program (as we get the option to overwrite a database series in the Store(db=xyz,o) command).

3. Graph legends: manually modified graph legends are not preserved (they revert to default) when one re-generates the graph using an existing version as ‘template’ for a new update. Eviews gives the option to set display names for series – one can display simple English legends rather than series names – but we have to manually ‘simplify’ legends when we use on-the-fly transformations [e.g. @pcy(@movav(series,3)) ].

Often, we don’t graph the raw series. If we change legends manually, this change is not maintained updating the graph using the old one as the ‘template’. Again, this problem only applies during automation of graph updates – it is not an issue in a one-time graph creation. And, this may have already changed in the Eviews 7 graphing scheme.

4. Immediately comes to mind a very similar problem with data-tables. After you interactively overwrite the label of a row, it reverts back to underlying formula following any modification. For this reason, I cannot replicate a data table I have prepared for country ABC for country XYZ. As soon as I change the data codes (e.g. a group member from, say, countryabc_gdp/series1 to countryxyz_gdp/series2), the labels are again the underlying formulas. Any chance of a ‘do not refresh labels’ option? An even better option could be to introduce ‘table templates’ just as we have graph templates. These would make some rows bold, fix the decimal places, shade certain rows, take 1 year change where appropriate, use the same specified labels for each row etc.

5. Data table: slightly more crucial: we have the option in data tables to apply transforms such as 1 year chg. A key transform is missing: 1 year chg (year-to-date %). When observing data tables in lower frequency than native (say, viewing monthly data in an annual workfile page), if data for only a part of the year is available, the convention is to calculate year-to-date % changes – i.e. average change over the same months of the previous year. Such a calculation is required when data are seasonally unadjusted. Instead, Eviews by default averages the entire previous year, then averages the limited data of the current year, then calculates 1-yr change.

6. More annotation possibilities on graphs would be great (e.g. arrows, circle an area, send drawn objects back, or bring them front). Gareth, you already wrote to me a year ago that something along these lines have been included in Eviews 7. One more thing would help as well: to be able to format the same series differently at different points in the graph; e.g. in a bar chart, positive readings green, negative readings red; or select just one bar and make it another color.

7. An ‘update’ button for graphs would be v useful. If this feature is made available, please take the smpl also from the graph when last generated (and not what the workfile smpl is currently set at). I believe that Eviews 7 will include updating graphs, but have no idea how they will update. The forum post http://forums.eviews.com/viewtopic.php?f=8&t=584 I believe is also looking for such a feature.

8. It would be useful if the Genr and Fetch commands could use an option to ignore NA values when reading series from a database. We often fill missing values or store forecasts in the ‘missing’ part of a series in the workfile. Right now, there appears to be no way, apart from meticulous smpl’s set for each series, to retain modifications in the workfile when the original series in the database has an ‘NA’. Setting the smpl’s using ‘isna’ appeared to create some problems, as far as I can remember, especially when downconverting frequency on-the-fly. Part-year data would not give an ‘NA’.

Some other options would also be very useful: for example, when using high frequency data in a lower frequency page, the down-conversion, I think, is only by global default. Series-by-series options are only available if we agree to first use a high-frequency page to fetch the series, then copy-paste/link it to the lower frequency page. We can then choose to use last-period values, ignore incomplete periods etc. These same options would be very welcome as options within a genr command. I say genr, not fetch, because we want to fetch the series and transform on-the-fly in one step. E.g. genr(options: ignore Nas, ignore years with incomplete quarters) abc_gdp_deflator=@pcy(database::abc_ngdp/database::abc_rgdp).

9. Now, the major ones. ‘Live’ and interactive graphs (as opposed to frozen objects) would be a huge step forward. They would update when series in the underlying database had new observations (like groups/formulas update), but more importantly they could be explored and modified interactively – a zoom slider (or drawn rectangle) may control how much history is shown, drag the axes to change scaling etc. A number of apps such as Weisang’s Flexpro, Datagraph for the Mac, and exploration interfaces such as Tibco’s Spotfire do these things and would be worth examining as models.

10. A fuzzy pattern recognition algorithm would be great. For a defined group of variables it would search for a past period when similar (not exact) patterns were made (over a defined period, say, a year) by each of the member variables.

11. Neural networks for general non-linear models and VARs.

12. Finally, I would personally welcome and prefer a second UI in Eviews in addition to the workfile/objects interface – a document interface (like Mathcad’s). On this one could put together ‘live’ and automatically updating graphs, tables and user-text in the form of, say, a monthly performance dash-board. An entire chart-book could be programmed and stored, along with user comments, notes on methodology etc., and would be always up-to-date with respect to data series in a main database.


Regards,
Tatha