Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.
|
|
|
|
Re: True independent variable: help needed
Posted:
Jun 20, 2006 5:35 AM
|
|
> I think you are confusing two issues. Bill H addressed the issue you
> raise in your 2nd paragraph--i.e., fixed versus random factors. (For
> those reading in sci.stat.edu, Bill's post appeared in sci.stat.consult
> only.)
>
> But in the 1st paragraph, you are talking about the distinction between
> variables that the researcher can actually manipulate (e.g., dose of
> drug given to each subject) and those that can only be observed and
> recorded (e.g., age, gender), and you got it backwards. Variables that
> can be manipulated by the researcher are true independent variables.
Thanks for your help, guys.
Bruce - you're absolutely right, I inverted the true / non-true
definitions in my first post. Sorry about that - I guess it confused
my question even more!
Bill - I think I understand what you mean about random factors and
their effect on variance. As it happens, I'm following a
psycholinguistic convention (avoiding Clark's language-as-fixed-effect
fallacy) where I perform analysis by participants (where participants
are treated as a random factor) and by materials (where stimuli are
treated as a random factor). Stimulus type is treated as a fixed
factor in this design - which I think corresponds to what you said!
What I still don't understand - and I'd really appreciate it if either
of you guys could explain - is how the "true independent variable"
status relates to random factors. For example, I could argue that I'm
manipulating my factor of stimulus type (real-words, non-words) in a
similar way to manipulating a factor like dosage (drug, placebo)...a
true IV. On the other hand, I could argue that I'm only observing the
level of each stimulus (e.g., CAT is a real word, GAT is not) in a
similar way to observing the gender of each participant...a non-true
IV.
In short, it seems to me that I'm sampling both participants (from the
population) and stimuli (from the set of 3-letter strings)... Since I
can't randomly assign each stimulus to the real-word or non-word
categories, does this mean that stimulus type is not a true IV? Am I
making sense??
Thanks again,
Elle.
|
|
|
|
