Significance values in MI on AntConc

[E.A.Hoskins]

Dear Professor Anthony,

I am currently completing my MSc, and am really valuing the software and your work in this area. It's brilliant to have such software available and to be able to contribute to the body of work in this field.

I just have a quick question regarding MI in Collocation that I was hoping you might be able to quickly answer. Is there a conventional statistical value of significance for MI (in AntConc or by convention)? Baker (2008: 101) states that some tools divide the measure of MI by log base 2 and that in this instance ‘any score higher than 3 is usually deemed to be indicative of a strong collocation’. Would that be correct of AntConc 3.5.6. If not, would you please be able to point me in the direction of where I can best locate the conventional significance value? 

Many thanks,

Elspeth

[Laurence Anthony]

Hi Elspeth,

MI is an effect size measure so there is no significance cutoff as with statistical measures. However, there are two ways to proceed. As you say, one way is to just use a score (e.g. above 3), which  is a common convention. Another way is to use a statistical measure (e.g. log likelihood) first, and then rank the values by MI. This is possible in AntConc, too.

I tend to use the second approach.

In theory you can set a 95% confidence interval on MI values, which is equivalent to p<0.05 but few currently do this. It's also not implemented in AntConc.

I hope that helps.

Laurence.

###############################################################
Laurence ANTHONY, Ph.D.
Professor of Applied Linguistics
Faculty of Science and Engineering
Waseda University
3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan
E-mail: antho...@gmail.com
WWW: http://www.laurenceanthony.net/
###############################################################

--
You received this message because you are subscribed to the Google Groups "AntConc-Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to antconc+unsubscribe@googlegroups.com.
To post to this group, send email to ant...@googlegroups.com.
Visit this group at https://groups.google.com/group/antconc.
For more options, visit https://groups.google.com/d/optout.

[E.A.Hoskins]

Hi Laurence,

Thanks for your reply, that's really helpful.

Best,

Elspeth

To unsubscribe from this group and stop receiving emails from it, send an email to antconc+u...@googlegroups.com.