The Gamer's Quarter

[ryan]

I know dhex has some experience with SPSS, but I'm still fairly new to it and am needing some assistance. This might be an odd place to ask, but I know a few people around here are into economics and statistics, and I'm really needing some help on this.

Here's the situation: I need to write a paper that shows a bunch of regression business, so we pick a simple data set and go with it. I chose number of accidents per hour in Louisiana.

Y = Number of Accidents (Dependent)
X1 = Time of Day (Independent)
X2 = Number of Licensed Drivers (Independent)

In SPSS I have columns for Y, X1, X2, Dummy (0 for Night and 1 for Day), y, x1_square, x2_square, x1y, x2y, x1x2.

From here I'm supposed to run a regression and use the information in the report to say something like 'if X1 is zero, then number of accidents to X2 would be ___' for the two variables. But I'm lost when it comes to the dummy. Do I make the dummy X3? I messed around with that, but I had no idea how to regress it. I need to get these figures, do a few formulas, and then finish the paragraph, then go on about strength of the formula, problems with the formula, suggestions for future research etc.

Heeellllppp!
_________________
Come to me, Mordel. We shall depart.

[Swimmy]

Yeah, the dummy should just be another variable. It should regress exactly the same way, if the dataset lines up properly. There should be a column that's just 1s and 0s, so assign that to a new variable, named x3 or whatever, and treat it exactly like the others. Ignore the interaction variables for now, just try to get the dummy to work. It should pump out a coefficient.

If the dataset doesn't have a column of 1s and 0s and they want you to fool around and make a dummy column, that's a whole different story.

I don't really know anything about SPSS though, I only know statistics, so if it's something weird with the program I can't help.

(Also I promise I'll have the Sam & Max review soon. I have to take the GREs today, but hopefully I'll have time afterward.)
_________________

"Ayn Rand fans are the old school version of Xenogears fanboys."
-seryogin

[ryan]

Thanks - I'm going to have to give this a go. The professor was no help on this. "Learn it on your own." Great, since I don't have $600 to buy the damn thing, I have 14 days to learn it, write a rough draft, write a final copy, and hope I have the figures right to finish either one with some degree of competency.

The GREs, from what I've heard, are easy. I took the GMAT, which is supposedly a bit harder. I forgot my exact score, but I remember it's in the range that 80% of the people get, so I can't imagine you'll have too much trouble with the GRE.


HERE is a screenshot of how my table looks. When I try to regress with X3 (where my icon is) as an independent variable, I get this error:

Warnings
For models with dependent variable Number of Accidents (Y), the following variables are constants or have missing correlations: Dummy Variable: Night (0) and Day (1) (X3). They will be deleted from the analysis.

As far as I can tell, I have it named as X3 and am regressing it as if it's a standard variable.
_________________
Come to me, Mordel. We shall depart.

[dhex]

the gre is hells of easy, btw.
_________________

[Swimmy]

Oh, huh. I really wish I knew that program. From here it looks like Y only has an entry every other event, and the dummy variable matches them all as 0. Which would explain the "constant" error message; it sees the variable as constantly 0 and therefore having no effect. I mean, I don't know, what's the deal with Time of Day, Dummy, and X1 all having twice as many entries? And the dummy variable alternating perfectly?
_________________

"Ayn Rand fans are the old school version of Xenogears fanboys."
-seryogin

[Swimmy]

Ok, I'm looking at it closer, and I think the data's weirded out. You have two times and two dummy variables for each year, but only one accident figure per year. Is that supposed to be split up somehow? Because as of now, it looks like it's saying there were never any daytime accidents, or there's no data for daytime accidents. Or am I just completely misreading the screen because I don't know the program?
_________________

"Ayn Rand fans are the old school version of Xenogears fanboys."
-seryogin

[ryan]

Y is the total number of accidents per year. X1 is the time of day the accidents occur (night is the top, day is the bottom). X2 is the number of licensed drivers for that year. Dummy_Variable, X3, is showing 0 and 1, indicating which time of day the accidents occurred. So when you read the first line, it says 52,793 accidents happened during that year (1996) with 16,872 having occurred at Night (0=night) and 35,921 during the day (1=day). This could very well be set up wrong. It's been years since I did this and he just sort of skimmed over it.

--

Do I need to separate Day and Night crashes? That seems to be the only way I won't get an error. I make X1 Day, X2 night, and X3 total number of licensed drivers.
_________________
Come to me, Mordel. We shall depart.