Justifying Your Prediction in Words

1. Write a descriptive lead sentence.
   
   
• "If the trend established from 1994-2006 persists, we expect there to be approximately 243 million cell phone subscribers in 2008."
  
  
2. Write supporting sentences that speak to your confidence in the predictive power of the model.
   You should comment on:
   1. How well the model matches the data.
      
      
   2. How far into the future or past you can comfortably forecast.
      
      
   3. Explain other reasons for your level of confidence (see "Other Factors that Impact Confidence" below).
      1. If you you are confident in your prediction, explain why you believe that the phenomenon will continue to follow the trend into the past and future.
      2. If you are not confident in your prediction, you should describe one more more potential problems with the prediction.
      3. If you are moderately confident in your prediction, you should explain the factors that both support and weaken your confidence.
         

• "We are confident in our prediction because the r-squared value of .9880 shows that the data has a strong linear relationship. The predicition is only 2 years into the future, even though technology changes rapidly, we don’t believe there will be any changes in the near future that would affect our prediction because the demand is growing steadily and cell phone contracts are becoming more affordable."
  
  

Other Factors that Impact Confidence

Below is a list of factors that either strengthen or weaken your confidence in a prediction. You should comment on any that apply in your supporting sentences.

1. Do you have a logical reason to believe that you are focusing on a localized trend? How far into the future or past can you project this trend? You should explain how this impacts your confidence in your prediciton.
   
   
• "I am moderately confident that the trend will be linear for the next two years, but the trend is likely to level-off over the next five years as the the number of cell phone subscribers is projected to equal the total population."
• "Projecting the trend back 70 years, predicts that the the average American lifespan was 0 years. Since this is impossible, the data must only be describing a localized linear trend driven by recent medical advances." It is inappropriate to project such a trend into the past without understanding the factors that drive it.
  
  
2. Do the data suggest that you are focusing on a localized trend? In this case, you may not have logical reasons to suspect that the trend will change, but the last few data points suggest this nonetheless. For examples, your model may predict a higher value when the data from the last several years were decreasing. Explain why you think the model will or will not still apply into the future.
   
   
3. If less than 7 data points were available, the model may not have good predictive power. If you are confident in the model, justify your reasons for trusting the model regardless. If limited data weakens your confidence, explain.
   
   
4. If you omitted one or more data points to improve the quality of model, state that you did this and rationalize why it was for you to do so. How does this affect your confidence?