Dear friends, I have a statistical question. Sometimes, if I compare boys to girls on a specific variable, the error bars (confidence interval of means) seem to overlap slightly. Still, when I run a t-test, I find statistically significant differences. The rule is clear: if the confidence intervals do not overlap, then there is statistically significant difference. But if they overlap slightly, we have to use a t-test to know for sure if the the two means differ significantly. The point is: is there a rule of thumb to say, for example, "if the overlap is less than 20% of the length of the standard error, then a t-test would give significant results?" thank you for your time P.S.1 is there an easy way to plot error bars in R? P.S.2 an interesting discussion about this - highly recommended to read it - can be found at http://scienceblogs.com/cognitivedaily/2007/03/ill_bet_you_dont_understand_er.php jason Dr. Iasonas Lamprianou Assistant Professor (Educational Research and Evaluation) Department of Education Sciences European University-Cyprus P.O. Box 22006 1516 Nicosia Cyprus Tel.: +357-22-713178 Fax: +357-22-590539 Honorary Research Fellow Department of Education The University of Manchester Oxford Road, Manchester M13 9PL, UK Tel. 0044 161 275 3485 iasonas.lamprianou at manchester.ac.uk
you can write a function for yourself using arrows:
#x,y: dataset as vectors
#xe,ye: errors per entry as vectors, if the errors are symmetric
arrbar<-function(x,y,xe,ye){
l<-length(x)
for (i in 1:l) {
arrows(x[i],y[i],x[i]-xe[i]/2,y[i],angle=90,length=0.05)
arrows(x[i],y[i],x[i]+xe[i]/2,y[i],angle=90,length=0.05)
arrows(x[i],y[i],x[i],y[i]-ye[i]/2,angle=90,length=0.05)
arrows(x[i],y[i],x[i],y[i]+ye[i]/2,angle=90,length=0.05)
}
}
Iasonas Lamprianou schrieb:> Dear friends,
> I have a statistical question. Sometimes, if I compare boys to girls on a
specific variable, the error bars (confidence interval of means) seem to overlap
slightly. Still, when I run a t-test, I find statistically significant
differences. The rule is clear: if the confidence intervals do not overlap, then
there is statistically significant difference. But if they overlap slightly, we
have to use a t-test to know for sure if the the two means differ significantly.
The point is: is there a rule of thumb to say, for example, "if the overlap
is less than 20% of the length of the standard error, then a t-test would give
significant results?"
>
> thank you for your time
>
> P.S.1 is there an easy way to plot error bars in R?
> P.S.2 an interesting discussion about this - highly recommended to read it
- can be found at
http://scienceblogs.com/cognitivedaily/2007/03/ill_bet_you_dont_understand_er.php
>
> jason
>
> Dr. Iasonas Lamprianou
>
>
> Assistant Professor (Educational Research and Evaluation)
> Department of Education Sciences
> European University-Cyprus
> P.O. Box 22006
> 1516 Nicosia
> Cyprus
> Tel.: +357-22-713178
> Fax: +357-22-590539
>
>
> Honorary Research Fellow
> Department of Education
> The University of Manchester
> Oxford Road, Manchester M13 9PL, UK
> Tel. 0044 161 275 3485
> iasonas.lamprianou at manchester.ac.uk
>
>
>
>
>
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
--
_______________________________
Dipl.-Phys. Markus Schmotz
Universit?t Konstanz
Fachbereich Physik, Lehrstuhl Leiderer
Postfach M 676
D-78457 Konstanz
Tel.: +49 7531 88 3803, Fax: 3127
Mail: markus.schmotz at uni-konstanz.de
Iasonas, In response to PS.1 try error.bars in the psych package. Bill At 1:04 AM -0700 3/30/10, Iasonas Lamprianou wrote:>Dear friends, >I have a statistical question. Sometimes, if I compare boys to girls >on a specific variable, the error bars (confidence interval of >means) seem to overlap slightly. Still, when I run a t-test, I find >statistically significant differences. The rule is clear: if the >confidence intervals do not overlap, then there is statistically >significant difference. But if they overlap slightly, we have to use >a t-test to know for sure if the the two means differ significantly. >The point is: is there a rule of thumb to say, for example, "if the >overlap is less than 20% of the length of the standard error, then a >t-test would give significant results?" > >thank you for your time > >P.S.1 is there an easy way to plot error bars in R? >P.S.2 an interesting discussion about this - highly recommended to >read it - can be found at >http://scienceblogs.com/cognitivedaily/2007/03/ill_bet_you_dont_understand_er.php > >jason > >Dr. Iasonas Lamprianou > > >Assistant Professor (Educational Research and Evaluation) >Department of Education Sciences >European University-Cyprus >P.O. Box 22006 >1516 Nicosia >Cyprus >Tel.: +357-22-713178 >Fax: +357-22-590539 > > >Honorary Research Fellow >Department of Education >The University of Manchester >Oxford Road, Manchester M13 9PL, UK >Tel. 0044 161 275 3485 >iasonas.lamprianou at manchester.ac.uk > > > > > > >______________________________________________ >R-help at r-project.org mailing list >https://stat.ethz.ch/mailman/listinfo/r-help >PLEASE do read the posting guide http://www.R-project.org/posting-guide.html >and provide commented, minimal, self-contained, reproducible code.-- William Revelle http://revelle.net/revelle.html 2815 Lakeside Court http://revelle.net/lakeside Evanston, Illinois It is 6 minutes to midnight http://www.thebulletin.org
thank you Dr. Iasonas Lamprianou Assistant Professor (Educational Research and Evaluation) Department of Education Sciences European University-Cyprus P.O. Box 22006 1516 Nicosia Cyprus Tel.: +357-22-713178 Fax: +357-22-590539 Honorary Research Fellow Department of Education The University of Manchester Oxford Road, Manchester M13 9PL, UK Tel. 0044 161 275 3485 iasonas.lamprianou at manchester.ac.uk --- On Tue, 30/3/10, William Revelle <lists at revelle.net> wrote:> From: William Revelle <lists at revelle.net> > Subject: Re: [R] error bars > To: "Iasonas Lamprianou" <lamprianou at yahoo.com>, r-help at r-project.org > Date: Tuesday, 30 March, 2010, 13:56 > Iasonas, > > In response to PS.1 > try error.bars in the psych package. > > Bill > > > > At 1:04 AM -0700 3/30/10, Iasonas Lamprianou wrote: > >Dear friends, > >I have a statistical question. Sometimes, if I compare > boys to girls > >on a specific variable, the error bars (confidence > interval of > >means) seem to overlap slightly. Still, when I run a > t-test, I find > >statistically significant differences. The rule is > clear: if the > >confidence intervals do not overlap, then there is > statistically > >significant difference. But if they overlap slightly, > we have to use > >a t-test to know for sure if the the two means differ > significantly. > >The point is: is there a rule of thumb to say, for > example, "if the > >overlap is less than 20% of the length of the standard > error, then a > >t-test would give significant results?" > > > >thank you for your time > > > >P.S.1 is there an easy way to plot error bars in R? > >P.S.2 an interesting discussion about this - highly > recommended to > >read it - can be found at > >http://scienceblogs.com/cognitivedaily/2007/03/ill_bet_you_dont_understand_er.php > > > >jason > > > >Dr. Iasonas Lamprianou > > > > > >Assistant Professor (Educational Research and > Evaluation) > >Department of Education Sciences > >European University-Cyprus > >P.O. Box 22006 > >1516 Nicosia > >Cyprus > >Tel.: +357-22-713178 > >Fax: +357-22-590539 > > > > > >Honorary Research Fellow > >Department of Education > >The University of Manchester > >Oxford Road, Manchester M13 9PL, UK > >Tel. 0044? 161 275 3485 > >iasonas.lamprianou at manchester.ac.uk > > > > > > > > > > > > > >______________________________________________ > >R-help at r-project.org > mailing list > >https://stat.ethz.ch/mailman/listinfo/r-help > >PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > >and provide commented, minimal, self-contained, > reproducible code. > > > -- > William Revelle??? ??? > ??? http://revelle.net/revelle.html > 2815 Lakeside Court??? ??? > ??? http://revelle.net/lakeside > Evanston, Illinois > It is 6 minutes to midnight??? http://www.thebulletin.org >