I have samples of 14 different groups. I have built LDA model (generated discrimination plot) using 2 principle components. On both the x-axis and y-axis i have an option to select (1 to 14). what does the number 1 to 14 indicate? Could these be the discriminant functions? I am confused a little as discriminant functions are always one less than the actual number of groups and in which case i should be having till 13? Also, just an additional information, when I select (1) on both x axis and y-axis samples are aligned in a diagonal manner. Please give a thought on this.
Hi Rito, can you please plot the graph you are mentioning? I've used PCA-LDA on Unscrambler but I don't recall this options to select, maybe I was using a different software version. Could it be the number of PCA components to be selected for the plots?
Hello Nicola, the image file has been attached. As you will see in the figure that 9 and 14 are on the x and y axis respectively. These numbers (1-14) are interchangeable.
I suspect that these are linear discriminant scores (LDS) rather than linear discriminant function (LDF). In the book "Statistics and chemometrics for analytical chemistry by James NM and Jane CM" the authors have shown an example on page 225-226 where LDS are generated for each group (3 groups and 4 variables-therefore 3 LDS was generated). The difference between LDS and LDF can be seen in the book " Multivariate data analysis by Joseph FH Jr, William CB, Barry JB, and Rolph EA" on page 231 where the author defines LDF as ' a variate of the independent variable selected for their discriminatory power used in the prediction of group membership. The predicted value of the discriminant function is the discriminant Z score. Thus I believe that the 'numbers' indicates linear discriminant scores. What is your view?
Rito have you used numbers to distinguish category variables???? These do not represent scores. Category variables are shown on the x and y-axis. Hope this helps.
I have used numbers to assign different categories, and not variables. Kindly email me any reference which could be helpful. [email protected] @Praveen kumar Yadav. Anyway thank you both for reaching out.
- Badges
- Science topic
- Similar topics
- Correction