I'm struggling to get my head round the attached general linear model output - see attached.
The figures in the value column (circled in red) are apparently slopes (in the cases of continuous variables therefore time) or fixed parameter values (in the cases of factor variables therefore age, month, origin).
What is a fixed parameter value and how is it calculated?
We can start with a simple example. For categorical variables (like "month" in your model), the mean value for month 1 is equal to your intercept (-72.9). The mean of month 2 is the beta parameter added to this intercept. Just think in terms of the actual equation that you're dealing with. Thus, month 2 is -72.9+(-2.64). Each factor level is coded internally as 0 or 1. So, for month3, the beta parameter for all months except month 3 would be multiplied by zero. Month3 beta would be multiplied by 1. For continuous variables (like age), you would multiply the beta parameter by the age you want to predict a value for. Of course, you need to specify some value (0,1, or an age, etc.) for all 14 terms in your model. You can convince yourself of everything by hand calculating a predicted value and then checking it by using the predict function.
The way you have it now, you can only see the significant differences between month 1 and months 2-12. You'll have to judge this via p-values. There are a number of ways to get this. As you know, the t-value is simply the beta divided by its standard error. Then, you can get the p-value by:
Reply With Quote
Posting Permissions