|
| |
Features
PATN v3 is the first commercial release of PATN for Windows™. PATN started
life in 1984 as a practical tool for research into 'pattern analysis', a field
established by Professor Bill Williams and Dr Godfrey Lance at CSIRO in Canberra
Australia. Experience over the last 20 years has gone into this new version.
The DOS version contained an extremely broad range of
modules and options. Experience suggested that 5% of the options were used 95% of the time. PATN was designed to
confront users with an array of options, to ensure that it was not used as
a ''black box'.
PATN (for Windows) v1.00 builds on the knowledge of thousands of analyses. PATN examines an imported dataset and derives a suite of
appropriate options. A 'standard' analysis can now be done in a few seconds by
pressing one button. The other major innovation in PATN v3+ is the animated,
interactive ordination plot display. This is
the window where most interpretation occurs.
Take a look some screenshots of PATN
v3 to get an idea on how it looks. PATN v3, like it's predecessor DOS PATN, contains four basic components-
 | Input and output of data and results. PATN can import data in
Excel™ or 'csv' format, from old DOS PATN (both .prm/.dat/.clb/.rlb and .arc
formats). PATN can also and input or export data tables, group compositions,
ordination coordinates, dendrogram ordering and all its statistics.
|
| Preparation or pre-processing. PATN contains a range of
transformation and standardization options that can be quickly applied to
selected rows and columns of the Data Table. 'Visible Statistics' such as
minima, means, maxima, values > 0, sum and missing values (user
selectable) are also displayed in margins of the Data Table. PATN can also
rapidly display (a button press)-
 | Histograms of one or more columns or rows of the Data Table |
| Scatter plots of two columns or rows of the Data Table
|
|
| Analysis. PATN contains the traditional mix of analysis utilities-
| Association. PATN options include Bray and Curtis, Minkowski
Series, Two-Step and Kulczynski. |
| Classification. PATN contains a hierarchical classification
module with options Nearest Neighbour, Furthest Neighbour, Flexible UPGMA
and Flexible WPGMA and a fast non-hierarchical module based on ALOC/B that
can handle millions of objects. |
| Ordination. PATN uses the SSH algorithm (semi-strong hybrid
multidimensional scaling). This algorithm can perform anything from pure
metric to pure non-metric multidimensional scaling. |
| Networks. PATN can display the display a minimum spanning tree on
the ordination plot to assist the evaluation of data structures.
|
|
| Analysis Evaluation. PATN contains a range of options that
are designed to help you evaluate the analysis-
| TWO-Way Table. PATN generates a coloured coded two-way table
of the classification of objects and variables. |
| Box and Whisker Plots. PATN's traditional box and whisker plots
can be sorted by Data Table order of variables, by Kruskal-Wallis values or
alphabetically. |
| ANOSIM across all groups or
all pair-wise groups. |
| Kruskal-Wallis. These values are used to estimate the utility of
variables to discriminate between a set of groups. The Kruskal-Wallis values
are a non-parametric equivalent to an f-ratio (using ranked values). |
| PCC (Principal Component
Correlation) runs multiple linear regression on selected variables fin the
Data Table and the ordination
coordinates. The result is a vector for each variable embedded in the ordination plot (with correlations calculated). |
| MCAO. This is PATN's 'Monte-Carlo Attributes in Ordination'
utility. PATN randomly assigns values of variables to objects in the
ordination plot and runs PCC a user-selected number of times (usually 100)
to evaluate the 'robustness' of the PCC results. |
| A-Priori groupings for objects can be imported into PATN and
used for diagnosis in the 3d ordination plot, box and whisker plots and
, ANOSIM. |
|
TABS are used in PATN to specify an analysis and evaluation scenario.
Once options have been selected, PATN will run the complete analysis.
Depending on the size of the dataset, a standard analysis a a reasonable system
will take less than a minute.
The key feature of PATN is the interactive 3d ordination plot. Just about
all aspects of an analysis can be visualized on the ordination plot. It is this
display where you will (if you're' smart) spend most of your time while
using PATN. Features of this display include-
| Rotate the plot in any direction automatically or manually |
| Meaningful object colours (or black and white symbols or labels or
user-defined colours if you wish) or even group colors (PATN's or
user-defined/a-priori) |
| Display group centroids or all objects (size of objects is user-defined) |
| Select objects to display labels and highlight in Data Table |
| Drag between objects to display association value |
| Drag between group centroids to display best Kruskal-Wallis values |
| Display minimum spanning tree |
| Highlight differences between PATN and a-priori groups |
| Display data vectors from PCC |
| Display a legend |
| Add text |
| Save plot as an image or even as a video with the plot rotated through 360° |
A few of the many other features of PATN V3
| An extensive Help file with a comprehensive browse sequence |
| Analysis can be done on either rows (objects) or columns (variables) of
the Data Table. |
| Row or columns can either be intrinsic (used in the analysis) or extrinsic
(used only in the evaluation). Columns or rows can be selected and dragged
around the Data Table for better placement or to make extrinsic (to the right
or under a thick red line in the table) or intrinsic. |
| Extensive right mouse button functions in context for each window |
| The base display is the Data Table. User-selectable statistical
values displayed of all or selected rows or columns in their own TABS. |
| Most results are also readily available in separate TABS from the Data
Table. |
| The 'Recipe' of analysis and evaluation steps are maintained in a time
stamped log file (similar to what DOS PATN used to use). This file can be
exported as a text file. |
| Dendrograms are scalable (you can even display a group dendrogram) and the
fusion history is still visible in a dual panel. |
| Association matrices can be displayed, printed, exported an imported. |
Specifications
PATN v3 requires the following minimum configuration-
| Pentium 200 with 128MB memory, 3mb disk space and an 800 * 600 colour display |
| Windows™ 98, 2000 or XP |
"I think you will like PATN V3! At least those who continue to struggle
with DOS PATN will appreciate it! In the same way that DOS PATN set the
standards when it was developed, I hope that PATN v3 will become a highly
effective and fun tool for data analysis. I know it has been a long time coming,
but I wanted to ensure that we had a significant and effective new
product that was robust (hence v3!)."
Lee Belbin |
