-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlinear_model_4_exercise.html
636 lines (548 loc) · 20.4 KB
/
linear_model_4_exercise.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
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
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8" />
<meta name="generator" content="pandoc" />
<meta http-equiv="X-UA-Compatible" content="IE=EDGE" />
<title>Exercises</title>
<script src="site_libs/header-attrs-2.26/header-attrs.js"></script>
<script src="site_libs/jquery-3.6.0/jquery-3.6.0.min.js"></script>
<meta name="viewport" content="width=device-width, initial-scale=1" />
<link href="site_libs/bootstrap-3.3.5/css/flatly.min.css" rel="stylesheet" />
<script src="site_libs/bootstrap-3.3.5/js/bootstrap.min.js"></script>
<script src="site_libs/bootstrap-3.3.5/shim/html5shiv.min.js"></script>
<script src="site_libs/bootstrap-3.3.5/shim/respond.min.js"></script>
<style>h1 {font-size: 34px;}
h1.title {font-size: 38px;}
h2 {font-size: 30px;}
h3 {font-size: 24px;}
h4 {font-size: 18px;}
h5 {font-size: 16px;}
h6 {font-size: 12px;}
code {color: inherit; background-color: rgba(0, 0, 0, 0.04);}
pre:not([class]) { background-color: white }</style>
<script src="site_libs/navigation-1.1/tabsets.js"></script>
<script src="site_libs/navigation-1.1/codefolding.js"></script>
<link href="site_libs/font-awesome-6.4.2/css/all.min.css" rel="stylesheet" />
<link href="site_libs/font-awesome-6.4.2/css/v4-shims.min.css" rel="stylesheet" />
<style type="text/css">
code{white-space: pre-wrap;}
span.smallcaps{font-variant: small-caps;}
span.underline{text-decoration: underline;}
div.column{display: inline-block; vertical-align: top; width: 50%;}
div.hanging-indent{margin-left: 1.5em; text-indent: -1.5em;}
ul.task-list{list-style: none;}
</style>
<style type = "text/css">
.main-container {
max-width: 940px;
margin-left: auto;
margin-right: auto;
}
img {
max-width:100%;
}
.tabbed-pane {
padding-top: 12px;
}
.html-widget {
margin-bottom: 20px;
}
button.code-folding-btn:focus {
outline: none;
}
summary {
display: list-item;
}
details > summary > p:only-child {
display: inline;
}
pre code {
padding: 0;
}
</style>
<style type="text/css">
.dropdown-submenu {
position: relative;
}
.dropdown-submenu>.dropdown-menu {
top: 0;
left: 100%;
margin-top: -6px;
margin-left: -1px;
border-radius: 0 6px 6px 6px;
}
.dropdown-submenu:hover>.dropdown-menu {
display: block;
}
.dropdown-submenu>a:after {
display: block;
content: " ";
float: right;
width: 0;
height: 0;
border-color: transparent;
border-style: solid;
border-width: 5px 0 5px 5px;
border-left-color: #cccccc;
margin-top: 5px;
margin-right: -10px;
}
.dropdown-submenu:hover>a:after {
border-left-color: #adb5bd;
}
.dropdown-submenu.pull-left {
float: none;
}
.dropdown-submenu.pull-left>.dropdown-menu {
left: -100%;
margin-left: 10px;
border-radius: 6px 0 6px 6px;
}
</style>
<script type="text/javascript">
// manage active state of menu based on current page
$(document).ready(function () {
// active menu anchor
href = window.location.pathname
href = href.substr(href.lastIndexOf('/') + 1)
if (href === "")
href = "index.html";
var menuAnchor = $('a[href="' + href + '"]');
// mark the anchor link active (and if it's in a dropdown, also mark that active)
var dropdown = menuAnchor.closest('li.dropdown');
if (window.bootstrap) { // Bootstrap 4+
menuAnchor.addClass('active');
dropdown.find('> .dropdown-toggle').addClass('active');
} else { // Bootstrap 3
menuAnchor.parent().addClass('active');
dropdown.addClass('active');
}
// Navbar adjustments
var navHeight = $(".navbar").first().height() + 15;
var style = document.createElement('style');
var pt = "padding-top: " + navHeight + "px; ";
var mt = "margin-top: -" + navHeight + "px; ";
var css = "";
// offset scroll position for anchor links (for fixed navbar)
for (var i = 1; i <= 6; i++) {
css += ".section h" + i + "{ " + pt + mt + "}\n";
}
style.innerHTML = "body {" + pt + "padding-bottom: 40px; }\n" + css;
document.head.appendChild(style);
});
</script>
<!-- tabsets -->
<style type="text/css">
.tabset-dropdown > .nav-tabs {
display: inline-table;
max-height: 500px;
min-height: 44px;
overflow-y: auto;
border: 1px solid #ddd;
border-radius: 4px;
}
.tabset-dropdown > .nav-tabs > li.active:before, .tabset-dropdown > .nav-tabs.nav-tabs-open:before {
content: "\e259";
font-family: 'Glyphicons Halflings';
display: inline-block;
padding: 10px;
border-right: 1px solid #ddd;
}
.tabset-dropdown > .nav-tabs.nav-tabs-open > li.active:before {
content: "\e258";
font-family: 'Glyphicons Halflings';
border: none;
}
.tabset-dropdown > .nav-tabs > li.active {
display: block;
}
.tabset-dropdown > .nav-tabs > li > a,
.tabset-dropdown > .nav-tabs > li > a:focus,
.tabset-dropdown > .nav-tabs > li > a:hover {
border: none;
display: inline-block;
border-radius: 4px;
background-color: transparent;
}
.tabset-dropdown > .nav-tabs.nav-tabs-open > li {
display: block;
float: none;
}
.tabset-dropdown > .nav-tabs > li {
display: none;
}
</style>
<!-- code folding -->
<style type="text/css">
.code-folding-btn { margin-bottom: 4px; }
</style>
</head>
<body>
<div class="container-fluid main-container">
<div class="navbar navbar-default navbar-fixed-top" role="navigation">
<div class="container">
<div class="navbar-header">
<button type="button" class="navbar-toggle collapsed" data-toggle="collapse" data-bs-toggle="collapse" data-target="#navbar" data-bs-target="#navbar">
<span class="icon-bar"></span>
<span class="icon-bar"></span>
<span class="icon-bar"></span>
</button>
<a class="navbar-brand" href="index.html">PGR-LM</a>
</div>
<div id="navbar" class="navbar-collapse collapse">
<ul class="nav navbar-nav">
<li>
<a href="index.html">
<span class="fa fa-home"></span>
Home
</a>
</li>
<li>
<a href="setup.html">
<span class="fa fa-cog"></span>
Setup
</a>
</li>
<li class="dropdown">
<a href="#" class="dropdown-toggle" data-toggle="dropdown" role="button" data-bs-toggle="dropdown" aria-expanded="false">
<span class="fa fa-book"></span>
R Book
<span class="caret"></span>
</a>
<ul class="dropdown-menu" role="menu">
<li>
<a href="https://intro2r.com">
<span class="fa fa-firefox"></span>
Web book
</a>
</li>
<li class="divider"></li>
<li>
<a href="https://github.com/alexd106/Rbook/raw/master/docs/Rbook.pdf">
<span class="fa fa-file-pdf"></span>
PDF book
</a>
</li>
</ul>
</li>
<li class="dropdown">
<a href="#" class="dropdown-toggle" data-toggle="dropdown" role="button" data-bs-toggle="dropdown" aria-expanded="false">
<span class="fa fa-book"></span>
Exercises
<span class="caret"></span>
</a>
<ul class="dropdown-menu" role="menu">
<li>
<a href="exercises.html">
<span class="fa fa-book"></span>
Exercises
</a>
</li>
<li class="divider"></li>
<li>
<a href="exercise_solutions.html">
<span class="fa fa-book"></span>
Exercise Solutions
</a>
</li>
</ul>
</li>
<li>
<a href="data.html">
<span class="fa fa-download"></span>
Data
</a>
</li>
<li class="dropdown">
<a href="#" class="dropdown-toggle" data-toggle="dropdown" role="button" data-bs-toggle="dropdown" aria-expanded="false">
<span class="fa fa-question-circle"></span>
Info
<span class="caret"></span>
</a>
<ul class="dropdown-menu" role="menu">
<li>
<a href="syllabus.html">
<span class="fa fa-graduation-cap"></span>
Syllabus
</a>
</li>
<li>
<a href="People.html">
<span class="fa fa-user-friends"></span>
People
</a>
</li>
<li class="divider"></li>
<li>
<a href="resources.html">
<span class="fa fa-book"></span>
Resources
</a>
</li>
<li>
<a href="https://forms.gle/8xYAqv19x8SSAdfUA">
<span class="fa fa-commenting"></span>
Feedback
</a>
</li>
<li>
<a href="People.html">
<span class="fa fa-envelope fa-lg"></span>
Contact
</a>
</li>
<li class="divider"></li>
<li>
<a href="https://github.com/alexd106/PGR-LM">
<span class="fa fa-github fa-lg"></span>
Source code
</a>
</li>
<li>
<a href="https://twitter.com/Scedacity">
<span class="fa fa-twitter fa-lg"></span>
Twitter
</a>
</li>
</ul>
</li>
</ul>
<ul class="nav navbar-nav navbar-right">
</ul>
</div><!--/.nav-collapse -->
</div><!--/.container -->
</div><!--/.navbar -->
<div id="header">
<div class="btn-group pull-right float-right">
<button type="button" class="btn btn-default btn-xs btn-secondary btn-sm dropdown-toggle" data-toggle="dropdown" data-bs-toggle="dropdown" aria-haspopup="true" aria-expanded="false"><span>Code</span> <span class="caret"></span></button>
<ul class="dropdown-menu dropdown-menu-right" style="min-width: 50px;">
<li><a id="rmd-show-all-code" href="#">Show All Code</a></li>
<li><a id="rmd-hide-all-code" href="#">Hide All Code</a></li>
</ul>
</div>
<h1 class="title toc-ignore">Exercises</h1>
</div>
<p> </p>
<div id="model-selection-with-the-loyn-data" class="section level2">
<h2>Model selection with the Loyn data</h2>
<p> </p>
<p>In the previous exercise you fitted a pre-conceived model which
included the main effects of the area of the forest patch
(<code>LOGAREA</code>), the grazing intensity (<code>FGRAZE</code>) and
the interaction between these two explanatory variables
(<code>FGRAZE:LOGAREA</code>). This was useful as a training exercise,
and might be a viable approach when analysing these data if an
experiment had been designed to test these effects only. However, if
other potentially important variables are not included in the model this
may lead to biased inferences (interpretation). Additionally, if the
goal of the analysis is to explore what models explain the data in a
parsimonious way (as opposed to formally testing hypotheses), we would
also want to include relevant additional explanatory variables.</p>
<p> </p>
<p>Here we revisit the previous loyn data analysis, and ask if a
‘better’ model for these data could be achieved by including additional
explanatory variables and by performing model selection. Because we
would like to test the significance of the interaction between
<code>LOGAREA</code>, and <code>FGRAZE</code>, whilst accounting for the
potential effects of other explanatory variables, we will also include
<code>LOGAREA</code>, <code>FGRAZE</code> and their interaction
(<code>FGRAZE:LOGAREA</code>) in the model as before. Including other
interaction terms between other variables may be reasonable, but we will
focus only on the <code>FGRAZE:LOGAREA</code> interaction as we have
relatively little information in this data set (67 observations). This
will hopefully avoid fitting an overly complex model which will estimate
many parameters for which we have very little data. This is a balance
you will all have to maintain with your own data and analyses (or better
still, perform a power analysis before you even collect your data). No
4-way interaction terms in your models please!</p>
<p> </p>
<p>It’s also important to note that we will assume that all the
explanatory variables were collected by the researchers because <em>they
believed</em> them to be biologically relevant for explaining bird
abundance (i.e. data were collected for a reason). Of course, this is
probably not your area of expertise but it is nevertheless a good idea
to pause and think what might be relevant or not-so relevant and why.
This highlights the importance of knowing your study organism / study
area and discussing research designs with colleagues and other experts
in the field before you collect your data. What you should try to avoid
is collecting heaps of data across many variables (just because you can)
and then expecting your statistical models to make sense of it for you.
As mentioned in the lecture, model selection is a relatively
controversial topic and should not be treated as a purely mechanical
process (chuck everything in and see what comes out). Your expertise
needs to be woven into this process otherwise you may end up with a
model that is implausible or not very useful (and all models need to be
useful!).</p>
<p> </p>
<p>1. Import the ‘loyn.txt’ data file into RStudio and assign it to a
variable called <code>loyn</code>. Here we will be using all the
explanatory variables to explain the variation in bird density. If
needed, remind yourself of your data exploration you conducted
previously. Do any of the remaining variables need transforming
(i.e. <code>AREA</code>, <code>DIST</code>, <code>LDIST</code>) or
converting to a factor type variable (i.e. <code>GRAZE</code>)? Add the
transformed variables to the <code>loyn</code> dataframe.</p>
<p> </p>
<p>2. Let’s start with a very quick graphical exploration of any
potential relationships between each explanatory variable (collinearity)
and also between our response and explanatory variables (what we’re
interested in). Create a pairs plot using the function
<code>pairs()</code>of your variables of interest. Hint: restrict the
plot to the variables you actually need. An effective way of doing this
is to store the names of the variables of interest in a vector
<code>VOI <- c("Var1", "Var2", ...)</code> and then use the naming
method for subsetting the data set <code>Mydata[, VOI]</code>. If you
feel like it, you can also add the correlations to the lower triangle of
the plot as you did previously (don’t forget to define the function
first).</p>
<p> </p>
<p>3. Now, let’s fit our maximal model. Start with a model of
<code>ABUND</code> and include all explanatory variables as main
effects. Also include the interaction <code>LOGAREA:FGRAZE</code> but no
other interaction terms as justified in the preamble above. Don’t forget
to include the transformed versions of the variables where appropriate
(but not the untransformed variables as well otherwise you will have
very strong collinearity between these variables!). Perhaps, call this
model <code>M1</code>.</p>
<p> </p>
<p>4. Have a look at the summary table of the model using the
<code>summary()</code> function. You’ll probably find this summary is
quite complicated with lots of parameter estimates (14) and P values
testing lots of hypotheses. Are all the P values less than our cut-off
of 0.05? If not, then this suggests that some form of model selection is
warranted to simplify our model.</p>
<p> </p>
<p>5. Let’s perform a first step in model selection using the
<code>drop1()</code> function and use an <em>F</em> test based model
selection approach. This will allow us to decide which explanatory
variables may be suitable for removal from the model. Remember to use
the <code>test = "F"</code> argument to perform <em>F</em> tests when
using <code>drop1()</code>. Which explanatory variable is the best
candidate for removal and why?</p>
<p>What hypothesis is being tested when we do this model selection
step?</p>
<p> </p>
<p>6. Update and refit your model and remove the least significant
explanatory variable (from above). Repeat single term deletions with
<code>drop1()</code> again using this updated model. You can update the
model by just fitting a new model without the appropriate explanatory
variable and assign it to a new name (<code>M2</code>). Alternatively
you can use the <code>update()</code> function instead (I show you how
to do this in the solutions).</p>
<p> </p>
<p>7. Again, update the model to remove the least significant
explanatory variable (from above) and repeat single term deletions with
<code>drop1()</code>.</p>
<p> </p>
<p>8. Once again, update the model to remove the least significant
explanatory variable (from above) and repeat single term deletions with
<code>drop1()</code>.</p>
<p> </p>
<p>9. And finally, update the model to remove the least significant
explanatory variable (from above) and repeat single term deletions with
<code>drop1()</code>.</p>
<p> </p>
<p>10. If all goes well, your final model should be
<code>lm(ABUND ~ LOGAREA + FGRAZE + LOGAREA:FGRAZE)</code> which you
encountered in the previous exercise. Also, you may have noticed that
the output from the <code>drop1()</code> function does not include the
main effects of <code>LOGAREA</code> or <code>FRGRAZE</code>. Can you
think why this might be the case?</p>
<p> </p>
<p>11. Now that you have your final model, you should go through your
model validation and model interpretation as usual. As we have already
completed this in the previous exercise I’ll leave it up to you to
decide whether you include it here (you should be able to just copy and
paste the code).</p>
<p>Please make sure you understand the biological interpretation of each
of the parameter estimates and the interpretation of the hypotheses you
are testing.</p>
<p> </p>
<p><strong>OPTIONAL questions</strong> if you have time / energy /
inclination!</p>
<p> </p>
<p>A1. If we weren’t aiming to directly test the effect of the
<code>LOGAREA:FGRAZE</code> interaction statistically (i.e. test this
specific hypothesis), we could use AIC to perform model selection.
Repeat the model selection you did above, but this time use the
<code>drop1()</code> function and perform model selection using AIC
instead. Don’t forget, if we want to perform model selection based on
AIC with the <code>drop1()</code> function we need to omit the
<code>test = "F"</code> argument)</p>
<p> </p>
<p>A2. Refit your model with the variable associated with the lowest AIC
value removed. Run <code>drop1()</code> again on your updated model.
Perhaps call this new model <code>M2.AIC</code>.</p>
<p> </p>
<p>A3. Refit your model with the variable associated with the lowest AIC
value removed and run <code>drop1()</code> again on your new model
(<code>M3.AIC</code>).</p>
<p> </p>
<p>A4. Repeat your model selection by removing the variable indicated by
the model with the lowest AIC.</p>
<p> </p>
<p>A5. Rinse and repeat as above.</p>
<p> </p>
<p>If all goes well, your final model should be
<code>lm(ABUND ~ LOGAREA + FGRAZE + LOGAREA:FGRAZE)</code>. This is the
same model you ended up with when using the <em>F</em> test based model
selection. This might not always be the case and generally speaking AIC
based model selection approaches tend to favour more complicated minimum
adequate models compared to <em>F</em> test based approaches.</p>
<p>We don’t need to re-validate or re-interpret the model, since we have
already done this previously.</p>
<p>I guess the next question is how to present your results from the
model selection process (using either <em>F</em> tests or AIC) in your
paper and/or thesis chapter. One approach which I quite like is to
construct a table which includes a description of all of our models and
associated summary statistics. Let’s do this for the AIC based model
selection but the same principles apply when using <em>F</em> tests
(although you will be presenting <em>F</em> statistics and P values
rather than AIC values).</p>
<p>Although you can use the output from the <code>drop1()</code> (and do
a bit more wrangling) let’s make it a little simpler by fitting all of
our models and then use the <code>AIC()</code> function to calculate the
AIC values for each model rather than <code>drop1()</code>. Details on
how to do this are given in the solutions to this exercise.</p>
<p> </p>
<p>End of the model selection exercise</p>
</div>
</div>
<script>
// add bootstrap table styles to pandoc tables
function bootstrapStylePandocTables() {
$('tr.odd').parent('tbody').parent('table').addClass('table table-condensed');
}
$(document).ready(function () {
bootstrapStylePandocTables();
});
</script>
<!-- tabsets -->
<script>
$(document).ready(function () {
window.buildTabsets("TOC");
});
$(document).ready(function () {
$('.tabset-dropdown > .nav-tabs > li').click(function () {
$(this).parent().toggleClass('nav-tabs-open');
});
});
</script>
<!-- code folding -->
<script>
$(document).ready(function () {
window.initializeCodeFolding("hide" === "show");
});
</script>
<!-- dynamically load mathjax for compatibility with self-contained -->
<script>
(function () {
var script = document.createElement("script");
script.type = "text/javascript";
script.src = "https://mathjax.rstudio.com/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML";
document.getElementsByTagName("head")[0].appendChild(script);
})();
</script>
</body>
</html>