The fitting parameters, as well as related statistics, are displayed in the Fitting Log. You can print this results or save as HTML page.
To obtain the best curve fitting performance, you should scale your data to order one if the data is extremely large or small.
Analysis tools allow the location of exact points on the curve fit as well as differentiation and integration.
FindGraph can sort the entire data set, using the either the independent variable (X) or the dependent variable (Y) as the key.
The transformation dialog allows one to apply a function to column X and column Y of data.
Exporting a file allows you to easily transfer data out of FindGraph, most often to a spreadsheet. The exported file will contain the data points (X,Y).
FindGraph's nonlinear least squares curve fitter is very flexible. It enables you to take almost full control of the fitting process and to select the best initial parameter values by plotting the curve before the actual fitting and comparing it to the data.
Main measure of the "goodness of fit" is "the standard error of the estimate". This quantity measures the spread of the points around the fitting function.
FindGraph provides tools for importing ASCII data files. If your data is copied to the Clipboard, you can also paste your data into a FindGraph serie.
Data set can be scaled, translated, sorted, the outliers removed, or edited by hand similarly to a spreadsheet.
Worksheet data (X,Y) can be exported to another application by creating an export file or by copying the data to the Clipboard.
FindGraph automatically routes results from most fitting operations, including graphics and residuals charts, to the Fitting Log. You can print this results or save as HTML page.
Before deciding to define your own function, make sure that the function you need is not already among the built-in functions. This will increase computational performance dramatically.
When fitting from "Non-linear fitting" page, all the parameters by default vary during the iterative procedure. If you want to fix certain parameters to particular values, you must Fix them.
If you are having a problem with FindGraph, the best way to solve the problem is to contact me and describe the problem. E-mail to: serg@uniphiz.com.
Often, there will be a requirement to display FindGraph's plots in another application such as a word processor or spreadsheet. To save the plot in the graphic format (bmp, gif, jpg etc.), let paste this plot to MS Paint.
FindGraph allows you to define your own nonlinear regression models. All you have to do is select <Formula> and type in the equation.
Use <Distribution> view mode to find and show density distribution picks. By default brightness and color are proportional to density. Use logarithmic scale, to increase contrast. It is useful for maximums searching.
In <Diagram> view mode points are grouped in cells. You see the summary number of points in cell. The means and trends (X,Y) are built on points of each cell.
Various data transformation techniques are available. You can transform data: translate, rotate, smooth, subtract baseline, differentiate, integrate.
FindGraph's nonlinear least squares curve fitter is very flexible. It enables you to take almost full control of the fitting process, to select the maximum number of iterations and initial step.
All nonlinear regression algorithms require initial guesses to commence the search for the optimum regression parameters. In the case of a poorly performing curve fit, you must set these initial guesses manually.
Processes producing sigmoidal ("S-shaped") growth curves are common in a wide variety of applications. These curves start at a fixed point and increase their growth rate monotonically to reach an inflection point. After this, the growth rate approaches a final value asymptotically.
For a non-linear regression, to increase computational performance, scale your data sets to the order of one.
FindGraph enables you to take almost full control of the fitting process by plotting the precalced fitting curve before the actual fitting and comparing it to the data.
Filters for digitally filtering data in the Fourier domain: low pass, high pass, band pass, band block, and threshold filters.
The Fast Fourier Transformation (FFT) band block filter allows you to eliminate noise within a specified frequency range.
Before even starting FindGraph's nonlinear fitting, you must determine if you actually need it. If you want to fit simple function, you may be better off trying linear regression, which is also supplied with FindGraph.
An indispensable part of any fitting procedure is a good choice of initial parameter values. FindGraph enables you to select good initial parameter values in the easiest possible way. Plot your theoretical function for any choice of parameters. Thus, you can easily compare your data with the theoretical function for any choice of parameter values.
When the fitting procedure does not converge, the only way that you can "make" is to try to start fitting from several different initial parameter values and see what happens. You should scale your data.
FindGraph allows you to export information to the other programs and files. Points can be saved to ASCII tables. Fitting curves, points, and commentaries can be saved to XML file.
FindGraph is COM server. This means, that plot can be copied and pasted (embedded) directly into word processors, Excel or drawing programs. To edit embedded plot, double click the mouse button on it.
FindGraph's nonlinear regression method is based on the Broyden-Fletcher-Goldfarb-Schanno algorithm. The Simplex minimization method is provided as well.
The program supports OLE automation and can be built-in to user software. See examples (MS VB, MS VC++).
With graphical editor you can easily add, move, copy and delete single points or a series of points, remove and duplicate curves. Unlimited Undo/Redo.
A common need is to condense the data by fitting it to a model in the form of a parametric equation. The fitting model should be chosen to reflect that law so that the parameters in the curve fit have physical interpretation and meaning.
With graphical editor you can remove false points and add desirable points interactively.
Visit the FindGraph web site for up-to-date information on FindGraph. Select any item on the Help:FindGraph on the Web submenu to launch your web browser and visit that site.
It is possible to add points interactively. Often it is simply dribble graph than digitize. The program has a full set of drawing tools.
Data can be entered by digitizing screen charts from screen or from background picture.
Once your data have been fit, you can further analyze XY data and fit results. Evaluation option includes function, derivatives, and table generation.
FindGrap reads simple ASCII data files and can ignore comments or text in the file. File import features enable the reading of a wide variety of files. Columns can be skipped.
With "Trace Data" window it is possible to find zeroes and extremes of the function and to pick peaks of datasets.
FindGraph supplies 10 generic fits, including Linear Regression, Logistic Functions, Fourier approximation, Neural Networks plus a library of over 120 industry-specific formulas.
For a polynomial regression, you must supply what degree of polynomial that you would like. Polynomials usually tend to give decent curve fits to almost any data set, but higher order polynomials tend to oscillate badly.
Use Fast Fourier Transformation (FFT) filter to eliminate noise corresponding to frequency components that are below a specified threshold level.
Neural network approximation is equally suited for advanced and inexperienced users. Use it, if your goal is simply to create a smooth curve for interpolation or simulation.
The program allows you to build parametric line graphs X(u,v), Y(u,v) with formulas you input. You can draw families of lines with a given step of parameter.
You can add plugin module (DLL) to include your equations into FindGraph. Example C source code for a 'user model' plugin is provided in FindGraph install package.
Different weighting methods are supported by FindGraph: Statistical, Instrumental, Direct.
Asymptotic models are characterized by a monotonic growth from some fixed value to an asymptote. These models are most common the engineering sciences.
It is possible to find the first or second order derivatives for the active data plot.
With graphical editor you can easily add and move text commentary, draw text commentary along curve. Unlimited Undo/Redo.
"Confidence Ellipse" illustrates a probability distribution on different directions. Its "Volume" is proportional to the square root of a generalized dispersion.
At the FindGraph library there are more than 200 built-in famous 2-D curves from the areas of elementary mathematics, transcendental functions, probability and statistics, optimization, and graphics.
