|
Name:
|
MartiniMan
-
|
Subject:
|
You do not understand models
|
Date:
|
4/29/2020 8:35:31 AM
|
|
They were indeed wrong and anyone who thinks otherwise does not understand models. We use simple and complex models in our business and the goal of all of them is to be predictive of the future. The only purpose of a model is to provide a projection of what will happen in a week, month or year and decisions are made today based on what the model says will happen down the road. That is their sole purpose. When a model says X will happen and it doesn't, the model is wrong. There is no other option....right or wrong.
Models are mathematical algorithms that include a variety of input parameters used to populate the model so it can provide a predictive result. Some input parameters are based on historical data and other input parameters are not based on data, they are based on assumptions because no data exists.
The question is why was the model wrong. There are many possible explanations or combinations of explanations. First, the model may not be properly designed to meet the predictive needs. Second, the data used to construct the model was flawed in some way (i.e., the early data doesn't represent future data for some reason). Third, the assumptions made when data doesn't exist were wrong. In the case of climate models they have been wrong because of all three reasons. In the case of the Wuhan virus models I really don't know what went wrong. But what I do know from our use of models is that as new data becomes available you do what is called recalibration. That means you modify the input parameters using either new data or different assumptions and then you run it and the prediction is compared to actual results as they come in. Rinse and repeat. So what you see is that as time goes on the model becomes more accurate in predicting the future. But that really defeats the puprose in the short term, right? A model that consistently fails to predict the future and has to be constantly recalibrated is basically worthless in informing decision makers.
Sorry for the scientific lecture but it if you don't use models every day like we do you will not have the background to understand how they are supposed to work. So the fact is the models were wrong because they failed to even remotely accurately predict the future.
|