|
0
|
1 <tool id="rcve1" name="Compute RCVE" version="1.0.0">
|
|
|
2 <description> </description>
|
|
|
3 <requirements>
|
|
|
4 <requirement type="package" version="2.11.0">R</requirement>
|
|
|
5 <requirement type="package" version="1.7.1">numpy</requirement>
|
|
|
6 <requirement type="package" version="1.0.3">rpy</requirement>
|
|
|
7 </requirements>
|
|
|
8 <command interpreter="python">
|
|
|
9 rcve.py
|
|
|
10 $input1
|
|
|
11 $response_col
|
|
|
12 $predictor_cols
|
|
|
13 $out_file1
|
|
|
14 1>/dev/null
|
|
|
15 </command>
|
|
|
16 <inputs>
|
|
|
17 <param format="tabular" name="input1" type="data" label="Select data" help="Dataset missing? See TIP below."/>
|
|
|
18 <param name="response_col" label="Response column (Y)" type="data_column" data_ref="input1" />
|
|
|
19 <param name="predictor_cols" label="Predictor columns (X)" type="data_column" data_ref="input1" multiple="true">
|
|
|
20 <validator type="no_options" message="Please select at least one column."/>
|
|
|
21 </param>
|
|
|
22 </inputs>
|
|
|
23 <outputs>
|
|
|
24 <data format="input" name="out_file1" metadata_source="input1" />
|
|
|
25 </outputs>
|
|
|
26 <requirements>
|
|
|
27 <requirement type="python-module">rpy</requirement>
|
|
|
28 </requirements>
|
|
|
29 <tests>
|
|
|
30 <!-- Test data with vlid values -->
|
|
|
31 <test>
|
|
|
32 <param name="input1" value="reg_inp.tab"/>
|
|
|
33 <param name="response_col" value="1"/>
|
|
|
34 <param name="predictor_cols" value="2,3,4"/>
|
|
|
35 <output name="out_file1" file="rcve_out.dat"/>
|
|
|
36 </test>
|
|
|
37
|
|
|
38 </tests>
|
|
|
39 <help>
|
|
|
40
|
|
|
41 .. class:: infomark
|
|
|
42
|
|
|
43 **TIP:** If your data is not TAB delimited, use *Edit Datasets->Convert characters*
|
|
|
44
|
|
|
45 -----
|
|
|
46
|
|
|
47 .. class:: infomark
|
|
|
48
|
|
|
49 **What it does**
|
|
|
50
|
|
|
51 This tool computes the RCVE (Relative Contribution to Variance) for all possible variable subsets using the following formula:
|
|
|
52
|
|
|
53 **RCVE(i) = [R-sq (full: 1,2,..,i..,p-1) - R-sq(without i: 1,2,...,p-1)] / R-sq (full: 1,2,..,i..,p-1)**,
|
|
|
54 which denotes the case where the 'i'th predictor is dropped.
|
|
|
55
|
|
|
56
|
|
|
57 In general,
|
|
|
58 **RCVE(X+) = [R-sq (full: {X,X+}) - R-sq(reduced: {X})] / R-sq (full: {X,X+})**,
|
|
|
59 where,
|
|
|
60
|
|
|
61 - {X,X+} denotes the set of all predictors,
|
|
|
62 - X+ is the set of predictors for which we compute RCVE (and therefore drop from the full model to obtain a reduced one),
|
|
|
63 - {X} is the set of the predictors that are left in the reduced model after excluding {X+}
|
|
|
64
|
|
|
65
|
|
|
66 The 4 columns in the output are described below:
|
|
|
67
|
|
|
68 - Column 1 (Model): denotes the variables present in the model ({X})
|
|
|
69 - Column 2 (R-sq): denotes the R-squared value corresponding to the model in Column 1
|
|
|
70 - Column 3 (RCVE_Terms): denotes the variable/s for which RCVE is computed ({X+}). These are the variables that are absent in the reduced model in Column 1. A '-' in this column indicates that the model in Column 1 is the Full model.
|
|
|
71 - Column 4 (RCVE): denotes the RCVE value corresponding to the variable/s in Column 3. A '-' in this column indicates that the model in Column 1 is the Full model.
|
|
|
72
|
|
|
73
|
|
|
74 </help>
|
|
|
75 </tool>
|
