This is really a time series regression question. But I do all such analysis in SAS so I put it here.
There are two primary issues in time series I deal with. One is autocorrelation (which influences on the standard error) and the other is assuming that lags of Y or an X influence Y (this would lead to bias not just problems with the SE I think).
One way to address first order autocorrelation (the most common type in social science analysis) is to specify an autoregressive model of order 1 to the residuals, this is then used in the estimation process.
In sas this is:
PROC AUTOREG DATA=somefilename PLOTS =(ALL);
MODEL Y X1 X2 X3/ NLAG=1 METHOD=ML;
RUN;
where Y is the dependent variable and X1 through X3 are various predictor variables.
Say that you think that Xt influences Yt+1 (Y at a later data than t). Would the code above address the issues or do you actually have to add a new variable to the model that is X a period before Y (I think that the above code is already doing this, but obviously this is unclear to me).
There are two primary issues in time series I deal with. One is autocorrelation (which influences on the standard error) and the other is assuming that lags of Y or an X influence Y (this would lead to bias not just problems with the SE I think).
One way to address first order autocorrelation (the most common type in social science analysis) is to specify an autoregressive model of order 1 to the residuals, this is then used in the estimation process.
In sas this is:
PROC AUTOREG DATA=somefilename PLOTS =(ALL);
MODEL Y X1 X2 X3/ NLAG=1 METHOD=ML;
RUN;
where Y is the dependent variable and X1 through X3 are various predictor variables.
Say that you think that Xt influences Yt+1 (Y at a later data than t). Would the code above address the issues or do you actually have to add a new variable to the model that is X a period before Y (I think that the above code is already doing this, but obviously this is unclear to me).