IV-DV methods

CSGnet Archive CSG2006
1

[From Bill Powers (2006.12.29.1030 MST)]
Rick Marken and I have been discussing IV-DV methods, and I take the
position that there is nothing wrong with them except how the results are
interpreted. I came up with the following, which others might want to use
in their own words (if they haven’t already done so). The
thought-experiment is the observation of a rat’s preferred location in a
cage (in a cold room) as a function of the setting, high or low, of an
infra-red heater close to the cage.

···

============================================================================
According to the standard biological model, IV affects the senses, and
the resulting sensory signals affect the actions, so it seems that the IV
is causing the DV via the sensory signals – the change in stimulus
temperature is causing a change in the preferred location. What is left
out of this is the fact that the change in location is preventing
the change in stimulus from having its assumed effect on the sensory
signals. If you measured the skin temperature directly instead of just
assuming that it changes according to the heater setting, you’d find that
it shows almost no relationship either to the heater setting or to the
location in the cage. And that blows the S-R model to smithereens. The
rat is as likely to be in one location in the cage as another for any
skin temperature in the observed range. That’s because the rat moves to
keep its skin temperature close to some preferred value.

All you have to do is reverse one arrow:

Standard model: IV ---------> sensors ---------->
DV

PCT model IV
---------> sensors <---------- DV

It’s that simple. Of course the problem then becomes that of explaining
how the DV can possibly change in just the right way, and that’s what the
control-system model is for.

=============================================================================

Best,

Bill P.

2

[From Rick Marken (2006.12.29.1430)]

Bill Powers (2006.12.29.1030 MST) --

Rick Marken and I have been discussing IV-DV methods, and I take the position that there is nothing wrong with them except how the results are interpreted. I came up with the following, which others might want to use in their own words (if they haven't already done so)...

All you have to do is reverse one arrow:

Standard model: IV ---------> sensors ----------> DV

PCT model IV ---------> sensors <---------- DV

It's that simple. Of course the problem then becomes that of explaining how the DV can possibly change in just the right way, and that's what the control-system model is for.

As I said in private, I really like this formulation. But now that I have thought about it a bit more, I think it is fair to say that there is "something wrong" with IV-DV methods if the system under study is a control system. What is wrong with these methods is that they don't tell you what sensed aspect of the IV is responsible for the relationship between IV and DV when you are dealing with a control system. So even if you know the PCT model of the IV-DV relationship (IV --> sensors <--DV), you don't know what the CV is that sits between IV and DV (in the form of "sensors" in your diagram). Of course, you would have to build a CV into the control system model that explains the observed relationship between IV and DV. But you would eventually have to test this model -- which represents a hypothesis about the CV -- using the non-IV-DV method that is appropriate to the study of control systems: the Test for the Controlled Variable (TCV).

So while I never said (in my "Revolution" paper) that IV-DV methodology is "wrong", I think it's fair to say that IV-DV methods are the wrong way to study the behavior of control systems, which is the basic argument of my paper.

Best

Rick

···

---
Richard S. Marken Consulting
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400