Partial ROC: This program generates the AUC (Area Under the Curve) specifying the threshold on Y-Axis, which is 1-omission in this case (Ref : Peterson et al 2007). Program developed by Narayani Barve, Biodiversity Institute, Lawrence, KS. INSTALLATION INSTRUCTIONS: The process of installing this tool is very simple. The zip / rar file contains 3 files. This file contains setup.xyz, PartialROC1.pdf and COMDLG32.ocx. Setup.xyz is a file which contains the program setup and PartialROC1.pdf file contains the instructions file for installing and executing the program, which is in fact the file you are reading right now. For installing the program follow the steps below. 1. First extract the zip / rar file in a directory on hard drive. 2. Change the extension of setup.xyz to setup.exe. 3. Run the setup.exe file for installing the program. The setup program is very intuitive. This setup program is very similar to any other windows application setup. Supply the information setup program asks. 4. After the program is installed, a program group known as Partial ROC will be created. The program could be accessed by selecting Partial ROC option from Partial ROC group. 5. When you run the file, if you get error about COMDLG32.ocx (as shown in following figure), then you can either download the file from internet or you can copy the COMDLG32.ocx in the folder where you installed the program.
After the program is installed successfully installed, you can run the program. Details about the input parameter value, formats are given below. The purpose of this program is to generate partial ROC-AUC ratios. The output of the program is the ratio of AUC at the given 1- omission threshold to the AUC at 50% (random). For more details about how the ratio calculation refer the paper “Peterson AT, Papes M, Soberon J (2008) Rethinking receiver operating characteristic analysis applications in ecological niche modeling. Ecol Model 213:63–72.” The results are stored in a text file.
Screen capture of the program. PARAMETER DETAILS : Presences file – This file contains the testing data points. The program expects the data in a specific format. The file type is comma delimited. Any file extension is allowed. First row in the file is the header. Actual data should start after the header. The file should contain only 2 columns. First column contains the unique identifier for each test point, starting from 1 to…n. Second column must contain the suitability value (GARP, Maxent, or any other modeling application) of each test point. In this program it is assumed that the higher the suitability value, the better is the suitability. So the data should be arranged accordingly. (Maxent and GARP differ in this sense: higher Maxent cumulative values approximate higher omission error values whereas higher GARP model agreement values mean better suitability values). Sample format of the file is shown below ID,suitability value 1,100 2,100 3,100 4,99 5,98 | | N 1
Area-dependent suitability File: Specify the filename along with complete path and file extension.The format of the file must be comma delimited. This file gives area or count of pixels per class of model prediction value. First line in the file must contain the header. Actual data should start from second line. Sample format of the file is shown below. suitability value,Count/Area 0,0 1,0 2,0 3,0 4,0 5,0 6,0.007 | | N, 23.23 Proportion of points in bootstrap (0 – 100%): Specify here the percentage of testing points which will be included in each of the random subsets of points generated to calculate the partial ROC. Sampling with replacement (# of repetitions): Specify the number of iterations, i.e. the number of subsets of points generated to calculate the AUC ratio. 1 - Omission Threshold – Specify threshold value for which the AUC ratio is to be generated. This represents omission error tolerance. For example, if set to 0.95, it means that the accepted omission error is 5%. For a complete AUC curve use 0. Otuput File – Specify the name of output file with path. This file contains the output in the following format. IterationNo,AUCatValue 95, AUCAt 0.5, Ratio 0,0.250207305827353,0.220719063263932,1.1336 1,0.250178911930953,0.220719063263932,1.1336 2,0.26964828733768,0.235406788092738,1.2216 3,0.242471527717497,0.214267696289829,1.131 | | MaxIterationno, value, value, value The values at 0th iteration are the AUCs for the total testing points. Actual testing points iterations start from 1 to the maximum iteration number specified by the user. Save iteration coordinates? You will see a check box before this question. If you check that box, the coordinates will be saved in a file. The name of the file should be specified
in the text box which is after the prompt. An error will pop up if the file already exists OR if the path given in the file is not correct. The purpose of saving these coordinates is to draw the AUC curves for future. Saving the iteration coordinates are optional.
Click on the “Generate” button so start generating bootstrapping ROC ratios. Program will display the stage of completion at the lower left corner of window. When you close the program it will display a window where you will see the reference details of the program. Refer this program in publications as below: “Narayani Barve (2008), Tool for Partial-ROC (Biodiversity Institute, Lawrence, KS), ver 1.0.”
Note: A software development is a process and is never complete until user find bugs, and gives suggestion. If you find any bug or error, please contact
[email protected] with details of the problem.