Question

Let "fit" summarize how far the in-sample observations are from the fitted model; e.g., fit=100 indicates worse fit than fit=50. Let "penalty" be increasing in the size of a model, specifically the number of parameters in the model, given the same sample size; e.g., given sample size 80, penalty is larger for a model with 10 parameters than for a model with 5 parameters. To prevent overfitting when choosing a forecasting model, you could

maximize (fit + penalty)

minimize (fit + penalty)

maximize (fit - penalty)

minimize (fit - penalty)

Answer 1

Let "fit" summarize how far the in-sample observations are from the fitted model; e.g., fit=100 indicates worse fit than fit=50. Let "penalty" be increasing in the size of a model, specifically the number of parameters in the model, given the same sample size; e.g., given sample size 80, penalty is larger for a model with 10 parameters than for a model with 5 parameters. To prevent overfitting when choosing a forecasting model, you could

maximize (fit + penalty)

minimize (fit + penalty)

maximize (fit - penalty)

minimize (fit - penalty)

Answer - The correct answer is OPTION B i-e MININIZE ( fit + Penality ) According to the problem FIT summarized the deviation of fitted model and sample observation and tell how far is the actual observation form the fitted model .In the given case first of all we have sample size , the penality is increasing the size of the model which will ultimately increases the fit .So In order to prevent over fitting we have to minimize both the penalty and the fit .By minimizing the fit we could reduce the deviation of actual observation from fitted line making r-squared value near 1 .And by minimizing penality we could be able to reduce the increasing model parameters.This will optimize the fit near 50 which is considered as good fit

HENCE MINIMIZE (FIT + PENALTY ) IS THE CORRECT ANSWER

HAVE A GOOD DAY !