[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: Concerning the margins command
"Nick Cox" <email@example.com>
RE: st: Concerning the margins command
Thu, 12 Nov 2009 13:06:37 -0000
I don't know when Roy started using Stata, but if complexity is the issue, then I still think that the old Stata graphics that finished with version 7 is way better than Excel graphics I have seen. And it's still bundled with Stata; it's just that you need an old manual. Nevertheless the new graphics really isn't that scary. Like most other things, you just pick up the bits you need as and when you need them.
I agree with Roy to this extent: life is short, we are all very busy and sanity depends on saying, some of the time: Other people appear to find that worthwhile, but I don't have the time and inclination to get into it. (My personal list includes following almost all sport, reading Proust or James Joyce, finding out what post-modernism means, etc.)
But two little stories:
1. I happen to be reviewing a paper right now in which the Excel-created graphics are so bad that they seriously get in the way of a proper analysis. Judging by what I have often seen in non-Stata presentations, whole groups of scientists are in the habit of doing their basic graphics in Excel and because they typically interact with other such scientists, they don't even know what they are missing. Why are these axes going through the middle of the data so that the ticks and labels are all mixed with the data points? Oh, that's what Excel does when there are negative values. But it doesn't have to be that way! Why are these zeros mixed up with the axis? It doesn't have to be that way!
2. I am happy to believe that if you know a lot of Excel you can do a lot better than the defaults. But from what I have seen the defaults are really lousy. The other day I created a graph for the first (!) time in Excel. (I decided, as a matter of military intelligence, that I needed to know a little more about the enemy.) I typed some numbers in columns A and B, rummaged around in the menu and found a scatter chart icon. Now, I can see why people like that side of things. But the result is so poorly designed that, even with low expectations, I was shocked:
- As far as I am concerned, my variables are A and B, but neither name appears on or off the graph.
- I get a legend with the text "Series 1". Who ordered that? One variable only is being plotted, so a legend is totally dispensable. And "Series 1" is a name that Excel is thrusting upon me. (Yes, I know a little more than I am saying.)
- There is a heavy grid of horizontal lines. I am not against grids, as I have written elsewhere, but this one is in your face.
Of course, I am biased and prejudiced and I have devoted, on a rough calculation, about a million times more effort in developing Stata skills than developing Excel skills.
But there is a bottom line for Roy and those so-minded: Get hold of Michael Mitchell's book and look through the examples. (Reading "Speaking Stata" in the Stata Journal might help a bit too.)
> The margins command is great. It does everything. It slices. It
> dices. It makes julienne fries. Clients come in with
> SAS and SPSS problems and leave with Stata margins output.
I am not saying this is good or bad but as a feedback in addition to
what Phil provided:
I never learned to use -xi- and am not really looking forward to
margins or factor variable. I will probably end up avoiding it as long
as I can. My guess is that it will turn out for me like Stata
graphics, which is great but greatly complicated. I have a pile of
do-files stashed somewhere but I just don't use them unless I really
and absolutely have to, which means I currently do most of my graphs
in Excel. It's clunky but good enough for most outlet and doesn't eat
up my time.
* For searches and help try: