waves/docs/reference/TestModelPerformance.html at develop · GoreLab/waves

History

364 lines (314 loc) · 18 KB

Raw

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

<!DOCTYPE html>

<head>

<title>Test the performance of spectral models — TestModelPerformance • waves</title>

reference values and reports model performance statistics" />

<!--[if lt IE 9]>

<![endif]-->

</head>

<span class="sr-only">Toggle navigation</span>

</button>

<a class="navbar-link" href="../index.html">waves</a>

</span>

</div>

<li>

</a>

</li>

<li>

<a href="../reference/index.html">Reference</a>

</li>

<li>

<a href="../news/index.html">Changelog</a>

</li>

</ul>

<li>

</a>

</li>

</ul>

</div>

</div>

</div>

</header>

<h1>Test the performance of spectral models</h1>

<small class="dont-index">Source: <a href='https://github.com/GoreLab/waves/blob/master/R/TestModelPerformance.R'><code>R/TestModelPerformance.R</code></a></small>

<div class="hidden name"><code>TestModelPerformance.Rd</code></div>

</div>

<p>Wrapper that trains models based spectral data to predict

reference values and reports model performance statistics</p>

</div>

<pre class="usage"><span class='fu'>TestModelPerformance</span><span class='op'>(</span>

<span class='va'>train.data</span>,

<span class='va'>num.iterations</span>,

test.data <span class='op'>=</span> <span class='cn'>NULL</span>,

preprocessing <span class='op'>=</span> <span class='cn'>TRUE</span>,

wavelengths <span class='op'>=</span> <span class='fl'>740</span><span class='op'>:</span><span class='fl'>1070</span>,

tune.length <span class='op'>=</span> <span class='fl'>50</span>,

model.method <span class='op'>=</span> <span class='st'>"pls"</span>,

output.summary <span class='op'>=</span> <span class='cn'>TRUE</span>,

rf.variable.importance <span class='op'>=</span> <span class='cn'>FALSE</span>,

stratified.sampling <span class='op'>=</span> <span class='cn'>TRUE</span>,

cv.scheme <span class='op'>=</span> <span class='cn'>NULL</span>,

trial1 <span class='op'>=</span> <span class='cn'>NULL</span>,

trial2 <span class='op'>=</span> <span class='cn'>NULL</span>,

trial3 <span class='op'>=</span> <span class='cn'>NULL</span>,

split.test <span class='op'>=</span> <span class='cn'>FALSE</span>,

verbose <span class='op'>=</span> <span class='cn'>TRUE</span>

<h2 class="hasAnchor" id="arguments"><a class="anchor" href="#arguments"></a>Arguments</h2>

<tr>

<th>train.data</th>

<td><p><code>data.frame</code> object of spectral data for input into a

spectral prediction model. First column contains unique identifiers, second

contains reference values, followed by spectral columns. Include no other

columns to right of spectra! Column names of spectra must start with "X"

and reference column must be named "reference".</p></td>

</tr>

<tr>

<th>num.iterations</th>

<td><p>Number of training iterations to perform</p></td>

</tr>

<tr>

<td><p><code>data.frame</code> with same specifications as <code>df</code>. Use

if specific test set is desired for hyperparameter tuning. If <code>NULL</code>,

function will automatically train with a stratified sample of 70%. Default

is <code>NULL</code>.</p></td>

</tr>

<tr>

<th>preprocessing</th>

<td><p>If <code>TRUE</code>, 12 preprocessing methods will be applied

and their performance analyzed. If <code>FALSE</code>, input data is analyzed as

is (raw). Default is <code>FALSE</code>.</p></td>

</tr>

<tr>

<th>wavelengths</th>

<td><p>List of wavelengths represented by each column in

<code>train.data</code></p></td>

</tr>

<tr>

<th>tune.length</th>

<td><p>Number delineating search space for tuning of the PLSR

hyperparameter <code>ncomp</code>. Default is 50.</p></td>

</tr>

<tr>

<th>model.method</th>

<td><p>Model type to use for training. Valid options include:</p><ul>

<li><p>"pls": Partial least squares regression (Default)</p></li>

<li><p>"rf": Random forest</p></li>

<li><p>"svmLinear": Support vector machine with linear

kernel</p></li>

<li><p>"svmRadial": Support vector machine with radial kernel</p></li>

</ul></td>

</tr>

<tr>

<th>output.summary</th>

<td><p>boolean that controls function output.</p><ul>

<li><p>If <code>TRUE</code>, a summary df will be output (1st row = means, 2nd row =

standard deviations). Default is <code>TRUE</code>.</p></li>

<li><p>If <code>FALSE</code>, entire

results data frame will be output</p></li>

</ul></td>

</tr>

<tr>

<th>rf.variable.importance</th>

<td><p>boolean that:</p><ul>

<li><p>If <code>TRUE</code>,

<code>model.method</code> must be set to "rf". Returns a list with a model

performance <code>data.frame</code> and a second <code>data.frame</code> with variable

importance values for each wavelength for each training iteration. If

<code>return.model</code> is also <code>TRUE</code>, returns list of three elements

with trained model first, model performance second, and variable importance

last. Dimensions are <code>nrow = num.iterations</code>, <code>ncol =

length(wavelengths)</code>.</p></li>

<li><p>If <code>FALSE</code>, no variable importance is

returned. Default is <code>FALSE</code>.</p></li>

</ul></td>

</tr>

<tr>

<th>stratified.sampling</th>

<td><p>If <code>TRUE</code>, training and test sets will be

selected using stratified random sampling. This term is only used if

<code>test.data == NULL</code>. Default is <code>TRUE</code>.</p></td>

</tr>

<tr>

<th>cv.scheme</th>

<td><p>A cross validation (CV) scheme from Jarquín et al., 2017.

Options for cv.scheme include:</p><ul>

<li><p>"CV1": untested lines in tested environments</p></li>

<li><p>"CV2": tested lines in tested environments</p></li>

<li><p>"CV0": tested lines in untested environments</p></li>

<li><p>"CV00": untested lines in untested environments</p></li>

</ul></td>

</tr>

<tr>

<th>trial1</th>

<td><p><code>data.frame</code> object that is for use only when

<code>cv.scheme</code> is provided. Contains the trial to be tested in subsequent

model training functions. The first column contains unique identifiers,

second contains genotypes, third contains reference values, followed by

spectral columns. Include no other columns to right of spectra! Column

names of spectra must start with "X", reference column must be named

"reference", and genotype column must be named "genotype".</p></td>

</tr>

<tr>

<th>trial2</th>

<td><p><code>data.frame</code> object that is for use only when

<code>cv.scheme</code> is provided. This data.frame contains a trial that has

overlapping genotypes with <code>trial1</code> but that were grown in a different

site/year (different environment). Formatting must be consistent with

<code>trial1</code>.</p></td>

</tr>

<tr>

<th>trial3</th>

<td><p><code>data.frame</code> object that is for use only when

<code>cv.scheme</code> is provided. This data.frame contains a trial that may or

may not contain genotypes that overlap with <code>trial1</code>. Formatting must

be consistent with <code>trial1</code>.</p></td>

</tr>

<tr>

<th>split.test</th>

<td><p>boolean that allows for a fixed training set and a split

test set. Example// train model on data from two breeding programs and a

stratified subset (70%) of a third and test on the remaining samples

(30%) of the third. If <code>FALSE</code>, the entire provided test set

<code>test.data</code> will remain as a testing set or if none is provided, 30%

of the provided <code>train.data</code> will be used for testing. Default is

<code>FALSE</code>.</p></td>

</tr>

<tr>

<th>verbose</th>

<td><p>If <code>TRUE</code>, the number of rows removed through filtering

will be printed to the console. Default is <code>TRUE</code>.</p></td>

</tr>

</table>

<h2 class="hasAnchor" id="value"><a class="anchor" href="#value"></a>Value</h2>

<p><code>data.frame</code> with model performance statistics in summary format

(2 rows, one with mean and one with standard deviation of all training

iterations) or in long format (number of rows = num.iterations).

<strong>Note</strong> if <code>preprocessing = TRUE</code>, only the first mean of

summary statistics for all iterations of training are provided for each

technique.

Included summary statistics:</p><ul>

<li><p>Tuned parameters depending on the model algorithm:</p><ul>

<li><p><strong>Best.n.comp</strong>, the best number of components</p></li>

<li><p><strong>Best.ntree</strong>, the best number of trees in an RF model</p></li>

<li><p><strong>Best.mtry</strong>, the best number of variables to include at every decision point in an RF model</p></li>

</ul></li>

<li><p><strong>RMSECV</strong>, the root mean squared error of cross-validation</p></li>

<li><p><strong>R2cv</strong>, the coefficient of multiple determination of cross-validation for PLSR models</p></li>

<li><p><strong>RMSEP</strong>, the root mean squared error of prediction</p></li>

<li><p><strong>R2p</strong>, the squared Pearson’s correlation between predicted and observed test set values</p></li>

<li><p><strong>RPD</strong>, the ratio of standard deviation of observed test set values to RMSEP</p></li>

<li><p><strong>RPIQ</strong>, the ratio of performance to interquartile difference</p></li>

<li><p><strong>CCC</strong>, the concordance correlation coefficient</p></li>

<li><p><strong>Bias</strong>, the average difference between the predicted and observed values</p></li>

<li><p><strong>SEP</strong>, the standard error of prediction</p></li>

<li><p><strong>R2sp</strong>, the squared Spearman’s rank correlation between predicted and observed test set values</p></li>

</ul>

<h2 class="hasAnchor" id="details"><a class="anchor" href="#details"></a>Details</h2>

<p>Calls <code><a href='DoPreprocessing.html'>DoPreprocessing</a></code>, <code><a href='FormatCV.html'>FormatCV</a></code>,

and <code><a href='TrainSpectralModel.html'>TrainSpectralModel</a></code> functions.</p>

<h2 class="hasAnchor" id="author"><a class="anchor" href="#author"></a>Author</h2>

<p>Jenna Hershberger <a href='mailto:[email protected]'>[email protected]</a></p>

<h2 class="hasAnchor" id="examples"><a class="anchor" href="#examples"></a>Examples</h2>

<pre class="examples"><div class='input'><span class='co'># \donttest{</span>

<span class='kw'><a href='https://rdrr.io/r/base/library.html'>library</a></span><span class='op'>(</span><span class='va'><a href='https://magrittr.tidyverse.org'>magrittr</a></span><span class='op'>)</span>

<span class='va'>ikeogu.2017</span> <span class='op'>%>%</span>

<span class='fu'>dplyr</span><span class='fu'>::</span><span class='fu'><a href='https://dplyr.tidyverse.org/reference/rename.html'>rename</a></span><span class='op'>(</span>reference <span class='op'>=</span> <span class='va'>DMC.oven</span><span class='op'>)</span> <span class='op'>%>%</span>

<span class='fu'>dplyr</span><span class='fu'>::</span><span class='fu'><a href='https://dplyr.tidyverse.org/reference/rename.html'>rename</a></span><span class='op'>(</span>unique.id <span class='op'>=</span> <span class='va'>sample.id</span><span class='op'>)</span> <span class='op'>%>%</span>

<span class='fu'>dplyr</span><span class='fu'>::</span><span class='fu'><a href='https://dplyr.tidyverse.org/reference/select.html'>select</a></span><span class='op'>(</span><span class='va'>unique.id</span>, <span class='va'>reference</span>, <span class='fu'>dplyr</span><span class='fu'>::</span><span class='fu'><a href='https://tidyselect.r-lib.org/reference/starts_with.html'>starts_with</a></span><span class='op'>(</span><span class='st'>"X"</span><span class='op'>)</span><span class='op'>)</span> <span class='op'>%>%</span>

<span class='fu'>TestModelPerformance</span><span class='op'>(</span>train.data <span class='op'>=</span> <span class='va'>.</span>,

tune.length <span class='op'>=</span> <span class='fl'>3</span>,

num.iterations <span class='op'>=</span> <span class='fl'>3</span>,

preprocessing <span class='op'>=</span> <span class='cn'>FALSE</span>,

wavelengths <span class='op'>=</span> <span class='fl'>350</span><span class='op'>:</span><span class='fl'>2500</span><span class='op'>)</span>

</div><div class='output co'>#> Preprocessing skipped.

#> Training model...</div><div class='output co'>#> Summary_type RMSEp R2p RPD RPIQ CCC Bias

#> 1 mean 2.1167852 0.75288809 1.95105646 2.5489867 0.851316240 0.3601674

#> 2 sd 0.2200048 0.02925815 0.09030978 0.2419144 0.009943066 0.2981776

#> SEP RMSEcv R2cv R2sp best.ncomp

#> 1 2.1387213 1.9679886 0.77622197 0.67944538 3

#> 2 0.2222847 0.1075555 0.01506289 0.07185676 NA</div><div class='input'><span class='co'># }</span>

</div></pre>

</div>

<h2 data-toc-skip>Contents</h2>

</nav>

</div>

<p>Developed by Jenna Hershberger, NSF BREAD IOS-1543958.</p>

</div>

<p>Site built with <a href="https://pkgdown.r-lib.org/">pkgdown</a> 1.6.1.</p>

</div>

</footer>

</div>

</body>

</html>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FilesExpand file tree

TestModelPerformance.html

Latest commit

History

TestModelPerformance.html

File metadata and controls