I have run a multinomial logistic model in SAS with 5 independant variables and I need to use the results from this model to make forecasts of use of care. I have used the predicted probabilities from the model and applied them to demographic forecasts per type of profile of client (age x gender) and now I need to use the exogenous variables (trend in disease, trend in unemployment, trend in handicap, etc) to forecast the use of care.

These exogenous variables are in my model also as independant variables and I have all coefficients and odds ratios for them. My question is how to apply the trend/forecast of these exogenous variables into my prediction?

How do I weight each of these exogenous variables in order to apply the growth of disease incidence, for example, into my forecast (demographic prediction)?

An example is how to apply disease incidence trend (ex. Alzheimer) into the forecast I have from the logistic regression? How do I calculate the weight of this variable?

  • 2
    $\begingroup$ If you know how the probabilities are related to the independent variables and parameters, and you have estimates for the parameters, and you have known or forecasted values for the independent variables, then you have the forecasted probabilities. Which part are you confused about? $\endgroup$ – The Laconic Sep 21 '18 at 12:32
  • $\begingroup$ Thank you. I have applied the predicted probabilities into the pre-existing demographical forecasts for the years 2015 until 2025. Now I wish to take this use the trends of the exogenous variables (ex.disease trends) and aplly them into the forecasts based on the demographics. Each exogenous variable has a coefficient and odds ratio in the model because they also served to construct the model. My question is how to construct weights to use their trends into the forecasts? Weights are needed because some variables like diseases have more impact in care use than others (ex. Income). $\endgroup$ – Nadine Sep 21 '18 at 17:53