README.md
In dhelkey/clustershard: Analyze and Cluster Compositional Pottery Data
clustershard
This package provides a GUI tool to analyze elemental compositional pottery data.
The tool simplifies the process of importing data, performing dimensionality reduction (through Principle Component Analysis), and clustering the observations (using k-means or Gaussian mixture modeling).
Overview
clustershard provides a GUI tool (clusterTool) for performing analysis on pottery shard data. Data is first imported from a CSV file and then the tool allows for different transformations. A clustering method (Gaussian Mixture Models or K-means) is then applied to a principle components decomposition of the transformed data. Finally, the tool provides methods for analyzing the results and exporting the resulting plots and data.
Getting Started
clustershard must first be loaded into R (requires the devtools package be installed) using:
devtools::install_github('dhelkey/dgrank', force = TRUE)
library('dgrank')
The GUI tool for clustering and analysis can then be run with:
clusterTool()
Tool Overview
-
Import data from a CSV file (format requirements).
-
Desired number of clusters.
-
Number of principle components on which the clustering is based.
-
Clustering method (Gaussian Mixture Models or K-means).
-
Data averaging. If TRUE, multiple observations of the same shard are combined into one reading by averaging element concentrations.
-
Data standardization. If TRUE, element concentration values are normalized to have mean 0 and standard deviation 1.
-
Data transformation. Options are 'None' for no transformation, 'Log' to take the natural log of each measurement, or 'LRA' which uses a Log-Ratio Analysis transformation (see Baxter & Freestone (2006) for details). This setting changes the data used to perform, and different settings will change the clustering results.
-
Visualization Transformation. Options are the same as above ('None', 'Log', 'LRA'). This transformation does NOT affect clustering results, but does affect the scale on which analysis plots are displayed.
-
Visualization data standardization. If true, element concentration values for plots are normalized to have mean 0 and standard deviation 1. This will not affect clustering results.
-
Export data along with clustering results.
-
Tabs for viewing the 5 different plots that clusterTool provides.
-
Drop down menu for selecting elements of interest.
-
Each plot provides a button to export the image to file.
Data Format
clusterTool assumes data is stored on a CSV file with one row for every measurement taken (there can be multiple rows for each specimen if it is measured multiple times).
The first row of the CSV should be a header row, with a SampleNo column, and columns with element names. The element names should start at at a fixed column, and continue to the end of the sheet(This can be set when running clusterTool() and defaults to 3). Element names can include numbers (for example the atomic mass of the element). Element names are compared with the periodic table and any unrecognized element names will throw an error. All columns to the right of the starting element column will be parsed as elements. Therefore, any superfluous columns need to be at the start of the CSV file, before the element starting column.
Identifying information about each measurement should be stored in a SampleNo column. This column should consist of text strings. The first element in the string should be an entry of the form "shardID-measurmentNum". Here, shardID is a unique identifier for each shard (e.g. D0506) and measurmentNum is an identifier for each measurement of the shard (e.g. 1a, 1b, 2a). Any extraneous information following this first dash separated value (e.g. a timestamp) is ignored by this package.
Example Acceptable Formats
By default, clusterTool assumes that numeric data starts at column 3 (from the left) and shard identifying information is in column 2. The DataFile column is ignored and numbers are removed when parsed by clusterTool:
DataFile | SampleNo | 7Li | 9Be | 23Na | 24Mg | 27Al | 29Si | 31P | 33S | 39K
---| ---| ---| ---| ---| ---| ---| ---| ---| ---| ---|
160824-A |D0506-1a 8/24/2016 1:28:10 PM |0.007235833 | 0.000278074 | 0.344942658 | 0.638880939 | 10.12715041 |43.54705309 | 0.12743092 | 0.087898704 | 1.069493988
160824-A | D0506-1b 8/24/2016 1:30:26 PM | 0.008610675 | 0.00052367 | 0.430605991 | 0.608618339 | 10.88459596 | 41.68874081 | 0.11672563 | 0.23707953 | 1.054835438
SampleNo | Li | Be | Na
---| ---| ---| ---|
D0501-a|0.007235833 | 0.000278074 | 0.344942658 | 0.638880939
D042-c| 0.008610675 | 0.00052367 | 0.430605991 | 0.608618339
Options
TODO
Author
- Daniel Helkey
All contributors listed in the package DESCRIPTION.
License
This project is licensed under the MIT License - see the LICENSE file for details
Acknowledgments
-
Judith Mauche proposed and funded the project, through [TODO agency or grant].
-
David Draper set up the collaboration and provided statistical consulting for the clustering tool.
-
R package constructed with extensive referral to to R Packages by Hadley Wickham
-
dhelkey/clustershard documentation built on May 9, 2019, 7:32 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
clustershard
This package provides a GUI tool to analyze elemental compositional pottery data. The tool simplifies the process of importing data, performing dimensionality reduction (through Principle Component Analysis), and clustering the observations (using k-means or Gaussian mixture modeling).
Overview
clustershard provides a GUI tool (clusterTool) for performing analysis on pottery shard data. Data is first imported from a CSV file and then the tool allows for different transformations. A clustering method (Gaussian Mixture Models or K-means) is then applied to a principle components decomposition of the transformed data. Finally, the tool provides methods for analyzing the results and exporting the resulting plots and data.
Getting Started
clustershard must first be loaded into R (requires the devtools package be installed) using:
devtools::install_github('dhelkey/dgrank', force = TRUE)
library('dgrank')
The GUI tool for clustering and analysis can then be run with:
clusterTool()
Tool Overview
-
Import data from a CSV file (format requirements).
-
Desired number of clusters.
-
Number of principle components on which the clustering is based.
-
Clustering method (Gaussian Mixture Models or K-means).
-
Data averaging. If TRUE, multiple observations of the same shard are combined into one reading by averaging element concentrations.
-
Data standardization. If TRUE, element concentration values are normalized to have mean 0 and standard deviation 1.
-
Data transformation. Options are 'None' for no transformation, 'Log' to take the natural log of each measurement, or 'LRA' which uses a Log-Ratio Analysis transformation (see Baxter & Freestone (2006) for details). This setting changes the data used to perform, and different settings will change the clustering results.
-
Visualization Transformation. Options are the same as above ('None', 'Log', 'LRA'). This transformation does NOT affect clustering results, but does affect the scale on which analysis plots are displayed.
-
Visualization data standardization. If true, element concentration values for plots are normalized to have mean 0 and standard deviation 1. This will not affect clustering results.
-
Export data along with clustering results.
-
Tabs for viewing the 5 different plots that clusterTool provides.
-
Drop down menu for selecting elements of interest.
-
Each plot provides a button to export the image to file.
Data Format
clusterTool assumes data is stored on a CSV file with one row for every measurement taken (there can be multiple rows for each specimen if it is measured multiple times).
The first row of the CSV should be a header row, with a SampleNo column, and columns with element names. The element names should start at at a fixed column, and continue to the end of the sheet(This can be set when running clusterTool() and defaults to 3). Element names can include numbers (for example the atomic mass of the element). Element names are compared with the periodic table and any unrecognized element names will throw an error. All columns to the right of the starting element column will be parsed as elements. Therefore, any superfluous columns need to be at the start of the CSV file, before the element starting column.
Identifying information about each measurement should be stored in a SampleNo column. This column should consist of text strings. The first element in the string should be an entry of the form "shardID-measurmentNum". Here, shardID is a unique identifier for each shard (e.g. D0506) and measurmentNum is an identifier for each measurement of the shard (e.g. 1a, 1b, 2a). Any extraneous information following this first dash separated value (e.g. a timestamp) is ignored by this package.
Example Acceptable Formats
By default, clusterTool assumes that numeric data starts at column 3 (from the left) and shard identifying information is in column 2. The DataFile column is ignored and numbers are removed when parsed by clusterTool:
DataFile | SampleNo | 7Li | 9Be | 23Na | 24Mg | 27Al | 29Si | 31P | 33S | 39K ---| ---| ---| ---| ---| ---| ---| ---| ---| ---| ---| 160824-A |D0506-1a 8/24/2016 1:28:10 PM |0.007235833 | 0.000278074 | 0.344942658 | 0.638880939 | 10.12715041 |43.54705309 | 0.12743092 | 0.087898704 | 1.069493988 160824-A | D0506-1b 8/24/2016 1:30:26 PM | 0.008610675 | 0.00052367 | 0.430605991 | 0.608618339 | 10.88459596 | 41.68874081 | 0.11672563 | 0.23707953 | 1.054835438
SampleNo | Li | Be | Na ---| ---| ---| ---| D0501-a|0.007235833 | 0.000278074 | 0.344942658 | 0.638880939 D042-c| 0.008610675 | 0.00052367 | 0.430605991 | 0.608618339
Options
TODO
Author
- Daniel Helkey
All contributors listed in the package DESCRIPTION.
License
This project is licensed under the MIT License - see the LICENSE file for details
Acknowledgments
-
Judith Mauche proposed and funded the project, through [TODO agency or grant].
-
David Draper set up the collaboration and provided statistical consulting for the clustering tool.
-
R package constructed with extensive referral to to R Packages by Hadley Wickham
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.