Help multiple regression

[MARGOT mh]

Dear list,

I want to run a multiple regression with this data file
http://www.statsci.org/data/oz/horses.html
I don understand with is the dependient variable. Does anyone know,how I can
do this?. I dont have  more information.

Thank you for your help.

Magi

_________________________________________________________________
Visita MSN Latino Noticias: Todo lo que pasa en el mundo y en tu paín, ¡en
tu idioma! http://latino.msn.com/noticias/

[Richard Ristow]

At 04:11 PM 12/26/2006, MARGOT MARIN wrote:

>I want to run a multiple regression with this data file
>http://www.statsci.org/data/oz/horses.html
>I don understand with is the dependient variable. Does anyone know,
>how I can
>do this?. I dont have  more information.

The first response - maybe people are too jaded to bother - is rather
sharp: "dependent variable" is NOT a property of the data, but of your
analytic question. That is THE most important point: analysis is not
about data, but about questions about the data. (Or, arguably, about
questions about the effects underlying the data.)

I'm reproducing, below, the variable list. from the site.

Exercise: Before you go farther, write down some things that you don't
know yet, that this data might help you learn.
..........................
Variable        Description

Position        Finishing position
Starters        Number of horses in race
Last            Finishing position in last race
Since           Days since last race
Number          Identifying number of horse in race
Carried         Weight carried
Weight          Handicap weight
Barrier         Barrier position at start of race
Distance        Length of race
Lengths         Number of lengths that horse finished from winner
Odds            Starting odds
Starts          Number of races previously started in
Age             Age of horse in years
Ratio           Proportion of wins in previous starts
..........................
Now, ONE use of regression is investigating possible causal effects:
what, is the result of what else? >NO REGRESSION ANALYSIS CAN PROVE
CAUSALITY< However, if there are *a priori* hypotheses of direction of
causality, a regression with the putative result variable as dependent,
and putative causal variables as independent, may give you grounds for
arguing for, or against, the hypothesis of causality.

One principle of causality is fixed: if the result and the cause are
separated in time, the result must be later in time than the cause.

In your data, this suggests ruling out all but the following as result
variables:
Position        Finishing position
Lengths         Number of lengths that horse finished from winner
Odds            Starting odds

Exercise: why? And list three other variables that could conceivably
be, at least partially, the results of other variables in your data.

* Multiple regression investigates the relationship between a set of
independent variables and a scale-level dependent variable.

Exercise: Which of the variables I've listed as possible dependent
variables, are most suitable as dependent variables in a multiple
regression? Why?

* Further, multiple regression investigates LINEAR relationships
between the dependent variable and scale-level, or dichotomous,
independent variables.

Exercise: Select independent variables, based on
. The above criteria
. A likely causal relationship (defend)

Exercise (advanced):
. Name variables that are otherwise suitable as independent variables,
but for which the relationship may well be non-linear.
..........................
I assume you're studying statistics, particularly multiple regression,
in course work. I recommend going over your answers to the exercises,
AND the principles I've laid out, with your instructors. (The
principles are, I think, generally sound, but in all cases open to
refinement; and in some cases, to argument.)

-Good luck,
  Richard Ristow

[Jeff-125]

At 12:14 PM 12/27/2006, you wrote:
>I assume you're studying statistics, particularly multiple regression,
>in course work. I recommend going over your answers to the exercises,
>AND the principles I've laid out, with your instructors. (The
>principles are, I think, generally sound, but in all cases open to
>refinement; and in some cases, to argument.)
>
>-Good luck,
>  Richard Ristow


...having taught intro stat to non-stat majors for about 4 years now,
this seems to me like the type of question asked by a student who
doesn't come to class and won't carefully examine the instructor's
instructions for completing assignments. I've spend hours and hours
carefully putting together detailed instructions for the completion
of basic work and still have a small minority of students who still
tell me that their dependent variable is gender (using survey data),
suicide (also using survey data), or the 150 students who
participated in the study. ...right afterward, they will explain that
the reason that they don't understand the assignment has nothing to
do with missing class over 50% of the time, but instead the ambiguity
of my questions. On the other hand, perhaps it is someone having a
good time with us.