How BestFit works
BestFit's goal is to find the distribution that best fits your input data. BestFit does not produce an absolute answer, it identifies a distribution that most likely produced your data.
For a given distribution, BestFit looks for the parameters of the function that optimize the goodness of fit, a measurement of the probability that the input data was produced by the given distribution. Always evaluate your BestFit results quantitatively and qualitatively, examining both the comparison graphs and statistics before using a result.
BestFit goes through the following steps when finding the best fit for your input data:
- For input sample data, parameters are estimated using maximum-likelihood estimators. For density and cumulative data, the method of least squares is used to minimize the distance between the input curve points and the theoretical function.
- Fitted distributions are ranked using one or more fit statistics, including Chi-square, Anderson-Darling, and Komolgorov-Smirnov.
BestFit gives you all the information you need to decide which fit is the best, and whether that fit is good enough to use. All results, including graphs, statistics and distribution functions, can easily be transferred to other programs for further analysis and presentation.
High resolution graphics are used to present the results from your BestFit run. Comparison, Difference, Probability-Probability (P-P) and Quantile-Quantile (Q?Q) graphs are all available. All graphs feature sliding delimiters to easily view specific probabilities or axis values. Want to compare a probability value for your input data and a fitted distribution? Simply slide the delimiter on the graph and see the calculated probabilities both on the graph and in the linked statistical report. Graph types and formatting can be quickly changed using toolbar icons, and rescaling is done by dragging axis limits directly on your graph. Change titles, colors, and more! All graphs can be displayed in Excel for further enhancement and hard copy. With BestFit, it's easy to clearly and effectively communicate complex results to others.
Detailed Statistics and Goodness of Fit Reports
BestFit generates full statistics and goodness of fit data for your fits in convenient tabbed reports linked to the active graph. Detailed statistics such as mean, variance, skewness, kurtosis and more are reported for the input data and resulting distribution. In addition, BestFit gives you access to full goodness of fit results for all three statistical tests (Chi-Square, A-D, and K-S). BestFit results are used with @RISK 4.5 Professional and Industrial for Excel, and can also be used with @RISK for Project and RISKOptimizer. With BestFit 4.5, all your simulations can take into account real-world data, giving you more accurate results!
Fit Summary Gives you Critical Values, Target Values and More!
In addition to the Statistics and Goodness of Fit tabs, BestFit can generate a comprehensive Fit Summary report. This report gives you, in convenient spreadsheet format, formulas, parameters, full statistics, and goodness of fit data for all fitted distribution functions. Plus, it displays Target Values and Critical Values for each distribution tested. Target Values tell you the probability of achieving a specific outcome, and Critical Values identify the criteria by which BestFit accepts or rejects a fit. These Critical Values can be used to determine how good a fit BestFit has found.
Customize Your Fits
Want to have more control over BestFit? You can tell BestFit which distributions you want it to try. BestFit also gives you full control over Chi-squared calculations, including equal interval binning, equal probability binning, and full custom binning. Furthermore, you can select predefined distributions with preset parameter values to fit. By giving you access to all these controls, BestFit allows you to tailor your fits to more precisely meet your needs. However, if you don't use all these features, BestFit will run fits automatically for you!
Get Up to Speed Fast!
BestFit comes with a complete on-line help system that uses an intuitive "How Do I?" interface. Simply type in what you want to do, or search the comprehensive index. Furthermore, the BestFit manual is written in plain, easy-to-understand English. If you would like further resources, Palisade carries a number of excellent titles on statistical fitting, including Risk Analysis: A Quantitative Guide, Simulation Modeling and Analysis and Statistical Distributions. And, Palisade offers ongoing seminars on how to use BestFit to analyze your data for use in @RISK, PrecisionTree, and other models! Visit www.palisade.com for the latest schedule.
Use With Other Palisade Software
BestFit was designed to work with our other DecisionTools Suite products - TopRank®, RISKview™, @RISK and PrecisionTree®. Start by using TopRank, the What-If Analysis add-in for Excel, to identify the critical values in your model. Next, in @RISK, the Risk Analysis and Simulation add-in for Excel, describe your uncertain values using probability distribution functions. Use RISKview to preview distributions or create one by simply drawing it and placing it directly in @RISK. Take real-world data and use BestFit to find the distribution which best fits your data, and use that function in @RISK to realistically represent uncertainty. Then run a thorough Monte Carlo simulation and analyze the results! Create a decision tree with PrecisionTree to accurately describe a decision, and use @RISK to perform a Monte Carlo simulation on your decision tree! With the DecisionTools Suite, you will be prepared to face the risks in any situation!
BestFit also works with RISKOptimizer™ and @RISK for Project. Use BestFit to find the best distribution for your data sets, and insert that distribution right in your Project or RISKOptimizer models! Then run a Monte Carlo simulation on your Project schedule with @RISK for Project, or run an optimization in Excel combined with Monte Carlo simulation with RISKOptimizer! One copy of BestFit is all you need to find accurate, reliable distributions to describe your uncertainty in any model!