Developing a multiple regression equation with repeated measures data

[LUM]

Hi,

I want to develop a multiple regression equation to predict pollutant concentrations (dependent variable) in runoff using six meteorological parameters (independent variables).

My experimental design involves me getting three observations under the same environmental/test conditions, i.e. I have three identical experiments that I measure pollutant concentrations in runoff under the same test conditions (the independent variables are the same for all the three experiments). I have collected data from 24 different experiment, so including the repeat measurements, I have in total 72 dependent variables to analyse.

Is there a test in SPSS that I can use to develop prediction models with repeated measurements? If so, is there a manual/paper that I use to help me develop these models?

I hope I supplied enough information for you.

Thanks in advance for helping me answer this question.

 

[staassis]

I do not think you are describing the problem correctly. This is not 24 eperiments but just 1. And your are dealing with 72 observations, not 72 dependent variables. My suspicion is that each measurement is the same factor but measured for different subjects under different conditions. Now, are you saying that you are measuring the same subject under 3 different conditions?

 

[hlsmith]

Would you call it longitudinal (panel) data?

 

[LUM]

Sorry I haven't explained myself properly. My experiments are designed so that I have three test subjects, the subjects are given the same treatments at the same time; therefore the independent variables are the same for the three subjects. The subjects produce runoff in this case which is analyzed for pollutant concentrations (DV). The results for the pollutant concentrations are different for the three subjects due to natural variance. The three subjects are exposed to 24 different treatments or experimental manipulations (different IV values) over time. This gave me 72 observations.

I would like to produce a prediction model that can predict the DV when different IV's are applied, but the model would have to take into account the repeated measures. I essential want a repeated measures ANOVA but using a regression model instead.

 

[staassis]

Thank you. The only estimation procedure in SPSS allowing for repeated measurements (and correlation between them) is Repeated Measures ANOVA. You invoke it by clicking

Analyze -----> General Linear Model -----> Repeated Measures.

You may ask the procedure to print the parameter estimates for each factor, which will essentially give you the desired predictive model. If you decide to prune that model by dropping certain (statistically insignificant) terms, it will have to be done by merging the relevant categories of the relevant factors in the data set.