&gt;<span style="font-family: Arial; font-size: 14.6666669845581px;">wold appreciate suggestions</span><div><font face="Arial"><span style="font-size: 14.6666669845581px;"><br></span></font></div><div><font face="Arial"><span style="font-size: 14.6666669845581px;">Frank Harrell: Regression Modeling Strategies. Springer.</span></font></div><div><font face="Arial"><span style="font-size: 14.6666669845581px;"><br></span></font></div><div><font face="Arial"><span style="font-size: 14.6666669845581px;">&gt;</span></font><span style="font-family: Arial; font-size: 14.6666669845581px;">I tried both methods</span></div><div><span style="font-family: Arial; font-size: 14.6666669845581px;"><br></span></div><div><span style="font-family: Arial; font-size: 14.6666669845581px;">Could you say what you did? Did you have a true model that the data were generated from, that the two methods were trying to find?</span></div><div><span style="font-family: Arial; font-size: 14.6666669845581px;"><br></span></div><div><font face="Arial"><span style="font-size: 14.6666669845581px;">&gt;</span></font><span style="font-family: Arial; font-size: 14.6666669845581px;">predictors significant at 95% level.</span></div><div><span style="font-family: Arial; font-size: 14.6666669845581px;"><br></span></div><div><span style="font-family: Arial; font-size: 14.6666669845581px;">I assume you mean 5%.</span></div><div><span style="font-family: Arial; font-size: 14.6666669845581px;"><br></span></div><div><span style="font-family: Arial; font-size: 14.6666669845581px;">Blaise F Egan</span></div><div><font face="Arial"><span style="font-size: 14.6666669845581px;"><br></span></font><blockquote style="margin-right: 0px; margin-left:15px;">----Original message----<br>From : <a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1502%26L%3DALLSTAT%26E%3Dquoted-printable%26P%3D1666734%26B%3D------%253D_Part_66109_5061871.1424290610695%26T%3Dtext%252Fhtml%3B%2520charset%3DUTF-8%26pending%3D" target="_parent" >[log in to unmask]</a><br>Date : 18/02/2015 - 19:04 (GMTST)<br>To : <a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1502%26L%3DALLSTAT%26E%3Dquoted-printable%26P%3D1666734%26B%3D------%253D_Part_66109_5061871.1424290610695%26T%3Dtext%252Fhtml%3B%2520charset%3DUTF-8%26pending%3D" target="_parent" >[log in to unmask]</a><br>Subject : multiple regression method sensitivity<br><br><meta http-equiv="Content-Type" content="text/html charset=us-ascii">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>multiple regression method sensitivity</title>

<font face="Arial"><span style="font-size:11pt">Stepwise regression is generally not recommended for multiple regression.<br>
Best subset is generally recommended instead? [sorry reference not to hand, wold appreciate suggestions]<br>
<br>
So being an empiricist I tried both methods on 52 sets of data with 21 potential predictors.<br>
I was surprised by results. There was a BIG difference in which method was most sensitive in terms of identifying predictors significant at 95% level.<br>
Most sensitive method identified a mean of 6 predictors, min=3, max =10<br>
Least sensitive method identified a mean of 4.25 predictors, min=2, max = 7<br>
<br>
Care to guess which method was most sensitive and explain why?<br>
<hr align="CENTER" size="3" width="95%">
Professor Diana Kornbrot<br>
email: : <a href="<a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1502%26L%3DALLSTAT%26E%3Dquoted-printable%26P%3D1666734%26B%3D------%253D_Part_66109_5061871.1424290610695%26T%3Dtext%252Fhtml%3B%2520charset%3DUTF-8%26pending%3D" target="_parent" >[log in to unmask]</a>"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1502%26L%3DALLSTAT%26E%3Dquoted-printable%26P%3D1666734%26B%3D------%253D_Part_66109_5061871.1424290610695%26T%3Dtext%252Fhtml%3B%2520charset%3DUTF-8%26pending%3D" target="_parent" >[log in to unmask]</a></a><br>
web: &nbsp;&nbsp;&nbsp;<a href="http://dianakornbrot.wordpress.com/">http://dianakornbrot.wordpress.com/</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<font color="#0000FF"><u><a href="http://go.herts.ac.uk/diana_kornbrot">http://go.herts.ac.uk/diana_kornbrot</a><br>
</u></font><b>Work <br>
</b>Department of Psychology<br>
School of Life and Medical Sciences<br>
University of Hertfordshire<br>
College Lane, Hatfield, Hertfordshire AL10 9AB, UK <br>
voice: &nbsp;&nbsp;+44 (0) 170 728 4626<br>
<b>Home<br>
</b>19 Elmhurst Avenue<br>
London N2 0LT, UK<br>
voice: &nbsp;&nbsp;+44 (0) 208 &nbsp;444 2081<br>
mobile: +44 (0) 740 318 1612<br>
</span></font>



You may leave the list at any time by sending the command
<p>
SIGNOFF allstat
</p><p>
to <a href="mailto:[log in to unmask]"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1502%26L%3DALLSTAT%26E%3Dquoted-printable%26P%3D1666734%26B%3D------%253D_Part_66109_5061871.1424290610695%26T%3Dtext%252Fhtml%3B%2520charset%3DUTF-8%26pending%3D" target="_parent" >[log in to unmask]</a></a>, leaving the subject line blank.
<br></p></blockquote><br><p></p></div>
You may leave the list at any time by sending the command
<p>
SIGNOFF allstat
</p><p>
to <a href="mailto:[log in to unmask]"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1502%26L%3DALLSTAT%26E%3Dquoted-printable%26P%3D1666734%26B%3D------%253D_Part_66109_5061871.1424290610695%26T%3Dtext%252Fhtml%3B%2520charset%3DUTF-8%26pending%3D" target="_parent" >[log in to unmask]</a></a>, leaving the subject line blank.