Roadmap Tutorial API's: XYGraph Axes etc Series Stats
Information in TDataSeries
TDataStats = record comment:String; error:boolean; count:integer; mean, median, mode, Total, SD, Skew, Kurtosis, lowquintile, highquintile, lowquartile, highquartile:double; end;
error is set to true if an error was encountered during analysis. Comment will contain a text description of the error. Some or all of the other values may be invalid if an error was encountered, they will be set to zero in this case
information in TCompStats
TCompStats= record comment:String; error:boolean; Rsquared, PearsonR, PRMin, PRMax, RSqMin, RSqMax, PValue, RegSlope, regintercept, RegSlope2, SlopeSD, IntSD: Double; end;
Error is set to true if an error was encountered during analysis. Comment will contain a text description of the error. Some or all of the other values may be invalid if an error was encountered, they will be set to zero in this case
Variable description
References: Numerical recipes in Pascal and Intuitive Biostatistics, H.M.Motulsky
Regression types and meaning of value stored in "r"
Type | identifier | value of r in series data |
none | rg_none | rubbish |
Least Squares | rg_linear | predicted value |
Passing Bablok | rg_passingBablok | predicted value |
Running Average | rg_runningAverage | average at point |
Spline | rg_spline | 2nd Differential at point |
Distance Weighted Least Squares (DWLS) | rg_DWLS | rubbish |
Quadratic | rg_Quadratic | predicted value |
References and comments on regression procedures:
Passing / Bablok Regression.
Bablok W. and Passing, H.: "Application of Statistical Procedures in Analytical instrument testing", J. Auto. Chem. v7, 1985, pp74-79. Suited for analytical comparisons - no assumption that x is without error.
Linear/Quadratic Regression. Devore, JL.: "Probability and Statistics for the engineering and Scientists", Duxbury Press, Belmont, Ca. 1990, pp460 - 461, or p516 - 517. I copied his quadratic correctly but it doesn't work well. Any body got a fix?
Running Average - applications in QA procedures. No reference
Distance Weighted Least Squares line fitting McLain DH.: "Drawing Contours from arbitrary data points." Comp. J., v17, 1973? pp 318 - 324. His discussion there is for 3d fitting - The weighting is discussed there. The least squares bit came off the top of my head, and I missed something, because it doesn't work quite as it should - see the comment in the main routine
Spline algorithm - adapted from the "Numerical Recipes in Pascal" toolkit