<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<style type="text/css" style="display:none;"><!-- P {margin-top:0;margin-bottom:0;} --></style>
</head>
<body dir="ltr">
<div id="divtagdefaultwrapper" style="font-size:12pt;color:#000000;font-family:Calibri,Arial,Helvetica,sans-serif;" dir="ltr">
<p><span style="font-size: 12pt;"><span style="font-family: Calibri, sans-serif; font-size: 14.6667px;">Joaquin ,</span><br>
</span></p>
<p><span style="font-size: 12pt;">The set of parameters which I have is stimulus parameter for tDCS (Current, PW, and some categorical variables). At the first
</span>level<span style="font-size: 12pt;">, I am planning to run a logistic regression model just to see if these
</span>parameters<span style="font-size: 12pt;"> would cause an impact ( measured by our method) or not. Then I will be using a stepwise GLM to model the actual level of output. </span><br>
</p>
<p>For the logistic model since the output is nominal ( TURE/FALSE), I cannot use the stepwise method. </p>
<p>any suggestions?</p>
<p>-Iman</p>
<br>
<br>
<div style="color: rgb(0, 0, 0);">
<div>
<hr tabindex="-1" style="display:inline-block; width:98%">
<div id="x_divRplyFwdMsg" dir="ltr"><font face="Calibri, sans-serif" color="#000000" style="font-size:11pt"><b>From:</b> Joaquin Rapela <rapela@ucsd.edu><br>
<b>Sent:</b> Wednesday, June 14, 2017 2:25 PM<br>
<b>To:</b> Iman Mohammad-Rezazadeh<br>
<b>Cc:</b> EEGLAB List; Anthony Zanesco<br>
<b>Subject:</b> Re: [Eeglablist] How to Interpret Insignificant Coefficients of Regression Analysis</font>
<div> </div>
</div>
</div>
<font size="2"><span style="font-size:10pt;">
<div class="PlainText">Dear Irman,<br>
<br>
The answer to your question depends on what you want to learn from the<br>
regression model.<br>
<br>
a. If I were interested in learning how a set of independent measures affect a<br>
dependent one, I would report both significant and insignificant coefficients.<br>
In this way the reader of my article would know the impact (and lack of impact)<br>
of each of the independent variables.<br>
<br>
b. If I ware interested in knowing what is the best model for my dependent<br>
variable, I would start with a model with a large number of independent<br>
variables, and their interactions, use some model selection criterion (e.g.,<br>
stepwise model selection with the Akaike information criterion), and only<br>
report the variables of the selected model.<br>
<br>
Cordially, Joaquin<br>
<br>
On Wed, Jun 14, 2017 at 06:10:28PM +0000, Iman Mohammad-Rezazadeh wrote:<br>
> Hi EEGLABers,<br>
> <br>
> This question is for mostly stats guru:<br>
> <br>
> If we develop a regression model (logistic, GLM, ...) and some of the<br>
> coefficient of the for the model is insignificant then how would<br>
> someone interpret that coefficient? should we through them away or<br>
> still keep them in the model? Should we consider their effect size as<br>
> the key parameter ?<br>
> <br>
> Thank you,<br>
> <br>
> Iman<br>
<br>
> _______________________________________________<br>
> Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" id="LPlnk935957" previewremoved="true">
http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
> To unsubscribe, send an empty email to eeglablist-unsubscribe@sccn.ucsd.edu<br>
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request@sccn.ucsd.edu<br>
<br>
<br>
-- <br>
Joaquin Rapela, PhD<br>
Swartz Center for Computational Neuroscience<br>
University of California San Diego<br>
9500 Gilman Drive,<br>
San Diego, CA 92093-0559<br>
tel: (858) 822-7547<br>
fax: (858) 822-7556<br>
<a href="http://sccn.ucsd.edu/~rapela/" id="LPlnk727307" previewremoved="true">http://sccn.ucsd.edu/~rapela/</a><br>
----------------------------------<br>
<br>
Assume a virtue if you have it not.<br>
<br>
Hamlet, Act 3, Scene 4<br>
William Shakespeare<br>
</div>
</span></font></div>
</div>
</body>
</html>