A dummy for your second wave would control for all factors that
influence your respondents uniformly. The measured variables than
assess the contributions of the variables above or below the common
trend. Much of the time, the observed time-varying predictors do not
fully account for temporal change. Putting the fixed-effect for time
is a good idea. If your observed variables do not fully account for
temporal change, omitting the dummy will bias your results. If the
observed variables do account for all of the temporal change, the
coefficient of the time duimmy will be reduced to insignificance. the
only cost is one degree of freedom used in the estimation. David
Greenberg, Sociology Department, New York University
On Fri, May 23, 2014 at 5:27 PM, Tino Nsenene <
[hidden email]> wrote:
> Dear all,
>
> I am planning to conduct a twolevel panel analysis with observations (level
> 1) nested within respondents (level 2).
>
> In total, there are six measurement occassions nested in 200 respondents. It
> seems to me that some authors include a variable (or several dummy
> variables) assessing the effect of "time" per se.
>
> I thinks such a predictor models the difference in the dependent variable
> over all measurement occassions. However, in my model I have substantive
> predcitor variables which should account for the overtime differences in the
> dependent variable. Hence, a priori I don't see any reason why I should
> additionally include "time" as covariate - also, I don't believe this
> question specifically refers to mixed models / multilevel models. The
> problem might also refer to repeatead anova type of models, to name just one
> additional example.
>
> Thus, I would be very grateful for any hints regardingthe inclusion of
> "time" as covariate... many thanks!
>
> Tino
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
>
[hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REF
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
