Predicted Values

[vivian1234]

Hi, I have some questions about the calculation of predicted values in runmlwin.

My model has two levels (level 1: individual; level 2: country). I was trying to fit a random slope model on my data with a continuous outcome. The continuous outcome ranges between 0 and 18. There are around 10 categorical and 5 continuous EVs at level 1 and there are 4 continuous level 2 EVs. The level 1 continuous EVs are all centered to mean. I have the random slope on one of the continuous EV (level 1).

When I was calculating the predicted values, I used the following syntax:

Code: Select all

gen ypred = [FP1]cons + [FP1]EV1*EV1 + u0 + u1* EV1
sort COUNTRY EV1
line ypred EV1, connect(a) 

As my outcome variable ranges between 0 and 18, I was expecting the predicted value will be in the same range. However, the predicted values ranges between -3 to 3. I have checked the mean and variance of the predicted value, they are neither mean zero or variance one. That's why I believe it is not due to the predicted value is being standardised.

Would anyone be able to give me some advice?

Thanks a lot.

Vivian

[ChrisCharlton]

In this prediction you are making the assumption that the parameters not included have a value of zero, which I suspect is not the case. If the prediction expression matched the model then you should get values closer to your original response variable.