### Aabel 統計分析軟體

** What's New?**

A Unique Solver for Kernel Density Estimate With AMISE Optimal Bandwidth Analyzer

• Efficient use of Kernel density estimate, which is

widely used in data mining and pattern recognition,

depends on computation of the optimal bandwidth of the kernel.

While AMISE (Asymptotic Mean Integrated Squared Error) optimal bandwidth can pro vide the best estimation, it involves extensive and time-consuming computation.

• Gigawiz has developed a unique hybrid solver for processing

the data and estimating the corresponding AMISE optimal

bandwidth by:

• Computing the univariate density derivatives for each data

point implementing the recommended methodology publish

ed by Raykar and Duraiswami(2005), followed by using the Sheather and

Jones method (1991) for the actual bandwidth estimation.

• The Gigawiz hybrid solver has been stress tested using

10000 data sets, including both real world and simulated

data.

**Gaussian Kernel Density Trace Charts**

Gaussian kernel density trace charts including:

• A graph type whose plotting involves processing the data

and computing the AMISE optimal bandwidth (using the

Gigawiz hybrid solver) while generating the resulting chart

on the fly

• A graph type with the rule of thumb bandwidth selectors

(i.e., Silverman's approach and percentage of sample range),

for exploratory visualization of the effect of bandwidth on

kernel density estimation

• Z-Score kernel density trace

• LOWESS (Robust Locally Weighted Regression) With

Weighted Confidence Interval (CI)

• LOWESS is an outlier resistant, robust locally weighted

regression and scatter smoothing, in which the fitted

value of xk is the value of a polynomial fit to the data

using weighted least squares, where the weight for

(xi, yi) is large if xi is close to xk and small if xi is far

from xk.

• The features and graph types include:

• Exploratory LOWESS Charts

• LOWESS Residual Dependence Graph

• Adding Weighted Confidence Interval (CI) to LOWESS Curves

• Using the LOWESS of Residuals as a Diagnostic Tool*

* The residual plot can provide information for choosing a

reasonable value for the local smoothing width).

**Weighted Mean**

• In weighted arithmetic mean, different weights are assigned

to different data points. In ordinary arithmetic mean, all

data points are assumed to have identical weight.

• The weighted mean has no analytic analogue of the standard

error of the mean for estimating the confidence limits.

**Violin Charts**

A violin chart comprises a combination of box plot and density trace;

violin chart types include:

• One-way violin

• Two-way violin

• Three-way violin

Bandwidth Selector Options:

• For interactive, exploratory visualization, you can choose one

of the following options:

• Silverman's approach using alpha

• Silverman's approach using IQR

• Percentage of sample range (for this option, you can obtain

the AMISE optimal bandwidth estimate (as well as the corres

ponding scaling factor)in theStats Analyzer and use the outp

ut for providing the user-defined scaling factor)

**XY and Multi-profile Scatter-Line Combination Charts**

XY Scatter-line combination charts*:

With data objects connected along the sorted values of Y-axis

With data objects connected along the sorted values of X-axis

* Why Combining XY Scatter and Line Chart Properties?

• In an XY scatter chart, the data points represent groups of

data objects that are plotted in the pipeline order (i.e., the

order they are stored in the worksheet)

• In a line chart, the data points represent variables plotted

along the sorted values of the Y- or X-axis.

• Combination graph types enable plotting a line chart in

which the data points represent groups of data objects.

Multi-profile scatter-line Combination Charts:

• Trends of multiple X-variables plotted across the sorted

values of a common Y-axis

• Trends of multiple Y-variables plotted across the sorted

values of a common X-axis

**Group-based n-Dimensional REE/Multi-Elements Diagrams**

• The newly added diagrams to specialized module of geochemic

al data provide flexibility for exploratory visualization of n-dime

nsional spidergrams generated from large data sets.

• Group-based n-dimensional REE or multi-elements diagrams

include:

• Group-based spidergrams differentiated by data object mark

ers and/or line attributes taken from the object marker color

properties

• Group-based spidergrams differentiated by color bands

**Open-GL Based 3D Grid Data Processor Utility**

• This Open-GL utility is designed for:

• Subtracting a grid Z-values from another (e.g., generating a

grid for an isopach map)

• Subtracting a grid Z-values from a user-defined XY plane

• Subtracting a user-defined XY plane from a grid Z-values

• Multiplying a grid Z-values by a constant, e.g., converting

the Z-values from feet to meter

• Re-sampling the displayed grid by user-defined number of

grid cells along the X and Y coordinates (e.g., for increasing

resolution of a grid for volume or area calculations)

**Open-GL Based 3D Grid Volume & Area Calculator Utility**

• This utility is designed for calculating a number of parameters

from a matrix by applying the calculations to:

• A user-defined polygon path comprising straight segments

• The total data coverage

**Enhancement of Utility Windows**

• The utility windows were re-designed to be modeless for

not blocking any other activities of the application.

**Optimization of Functions Interacting With Mac OS X Subsystems**

• The main enhancements include:

• Making the application more resilient to design errors of the

Apple HFS+ file system

• Workaround for enhancing the graphic performance that is

affected by low-performing parts of the Quartz Graphic Syste

m