tag:blogger.com,1999:blog-590043897961646114.post26014630312250367..comments2024-02-06T02:06:06.364-08:00Comments on Engaging Market Research: Archetypal AnalysisUnknownnoreply@blogger.comBlogger6125tag:blogger.com,1999:blog-590043897961646114.post-14205769609173704482012-10-01T09:40:43.250-07:002012-10-01T09:40:43.250-07:00No problem, thanks for the prompt correction. Chee...No problem, thanks for the prompt correction. Cheers!Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-590043897961646114.post-24838666460369003842012-09-28T13:37:25.182-07:002012-09-28T13:37:25.182-07:00So, it seems that some earlier code was inserted a...So, it seems that some earlier code was inserted accidentally. Thanks for catching this. Try the following which uses kcl_3$cluster to label the k-means from 1 thru 3 and uses the pca$var$coord from FactoMineR to label the arrows.<br /><br />#plots K-means<br />plot(pca$ind$coord[,1:2], type="n", xlim=c(-3.9,5.8), ylim=c(-3.9,5.8))<br />text(pca$ind$coord[,1:2], col=kcl_3$cluster, labels=kcl_3$cluster)<br />arrows(0, 0, 7*pca$var$coord[,1], 7*pca$var$coord[,2], col = "chocolate", angle = 15, length = 0.1)<br />text(7*pca$var$coord[,1], 7*pca$var$coord[,2], labels=1:10)<br /><br />#plots archetypes<br />plot(pca$ind$coord[,1:2], type="n", xlim=c(-3.9,5.8), ylim=c(-3.9,5.8))<br />text(pca$ind$coord[,1:2], col=aa_3_cluster, labels=aa_3_cluster)<br />arrows(0, 0, 7*pca$var$coord[,1], 7*pca$var$coord[,2], col = "chocolate", angle = 15, length = 0.1)<br />text(7*pca$var$coord[,1], 7*pca$var$coord[,2], labels=1:10)Joel Cadwellhttps://www.blogger.com/profile/14946447393733294251noreply@blogger.comtag:blogger.com,1999:blog-590043897961646114.post-35347149596203111762012-09-28T10:14:03.777-07:002012-09-28T10:14:03.777-07:00> text(7*axes$loadings[,1], 7*axes$loadings[,2]...> text(7*axes$loadings[,1], 7*axes$loadings[,2], labels=1:10)<br /><br />"Error in text(7 * axes$loadings[, 1], 7 * axes$loadings[, 2], labels = 1:10) : <br /> object 'axes' not found"<br /><br />I appreciate this blog post; please consider running your example code after clearing your workspace to ensure it's actually complete. Thanks.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-590043897961646114.post-38535558709621903702012-09-28T08:21:27.056-07:002012-09-28T08:21:27.056-07:00> text(pca$ind$coord[,1:2], col=kcl_3_labels, l...> text(pca$ind$coord[,1:2], col=kcl_3_labels, labels=kcl_3_labels)<br /><br />"Error in as.graphicsAnnot(labels) : object 'kcl_3_labels' not found"<br /><br />Presumably there is missing code?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-590043897961646114.post-59982259120794621682012-08-03T13:06:29.411-07:002012-08-03T13:06:29.411-07:00So I would conclude from your comment that you wou...So I would conclude from your comment that you would not find an archetypal analysis satisfactory if it did not uncover your “compromise” segment. It is a good point. Unlike supervised learning tasks where we have a criterion against which we can judge the “goodness” of our solutions, clustering is left struggling to find a standard for evaluating its procedures. A 2009 NIPS workshop on “Clustering: Science or Art?” (available at videolectures.net) suggested we adopt the application-specific criteria that a good clustering is one that is useful. Of course, usefulness must be defined, and you have just supplied one definition – it must be able to identify segments making attribute trade-offs. I believe that archetypal analysis can find such types if you include the correct measures. That is, you have told us that this is a segment focusing on versatility. By including versatile among our input measures, we ought to be able to find this group who values versatility above all else. Moreover, your question has shown how important it is to have substantive knowledge in order to select the right measures. The data analyst need to work closely with the substantive expert to understand what types are likely to be found and to be certain we have included the measures needed to identify each type.Joel Cadwellhttps://www.blogger.com/profile/14946447393733294251noreply@blogger.comtag:blogger.com,1999:blog-590043897961646114.post-32900940783330723902012-08-03T09:29:00.301-07:002012-08-03T09:29:00.301-07:00This is a really interesting read. I understand th...This is a really interesting read. I understand that archetypal analysis is capable of identifying "extremes" that are in essence some combination of multiple other extreme types. (IE. Using a basketball analogy, high shooting percentage with low frequency of shots taken as one type). My question is whether it is also suited for identifying, as an "extreme," a player that was about average in every category. In your 50-50 player height example, we are only concerned with identifying how short or tall the player is. I understand & agree with your point that "categorization gravitates toward black and white, premium and economy, self and full service, big and small businesses, ect." <br />But, does this research imply that there is no significant "segment" in certain consumer markets that specifically looks for a more "jack-of-all-trades" type of product? Paper towels is not the best example, but could there be a segment of the population that specifically looks for a versatile towel (medium thickness, medium size, ect) because they know they have to contend with various spill types in a given day. Such consumers may not have a singular focus such as saving money or having the ability to absorb the most moisture possible.Thomas Lofaronoreply@blogger.com