1
$\begingroup$

Basically, I have following LDA output (below). My question is how do I actually interpret it ? I could not find much resources on the web. Especially, what does "Group Means" really mean here ? I was expecting the group means to be the mean / centroids of the K-classes. Moreover, I am puzzled by the 2 terms as well, i.e. "coefficients of linear discriminants" and "proportion of trace"

Prior probabilities of groups:
         1          2          3          4          5          6          7          8          9         10 
0.09090909 0.09090909 0.09090909 0.09090909 0.09090909 0.09090909 0.09090909 0.09090909 0.09090909 0.09090909 
        11 
0.09090909 

Group means:
         x.1       x.2        x.3       x.4         x.5       x.6         x.7        x.8         x.9        x.10
1  -3.359563 0.0629375 -0.2940625 1.2033333  0.38747917 1.2218958  0.09637500 0.03710417 -0.62435417 -0.16162500
2  -2.708875 0.4906042 -0.5802292 0.8135000  0.20193750 1.0634792 -0.19091667 0.37381250 -0.51595833  0.08060417
3  -2.440250 0.7748750 -0.7983958 0.8086667  0.04245833 0.5692500 -0.28006250 0.20495833 -0.47827083  0.18187500
4  -2.226604 1.5258333 -0.8744375 0.4221458 -0.37131250 0.2483542 -0.01895833 0.10714583 -0.32627083 -0.05375000
5  -2.756312 2.2759583 -0.4657292 0.2253125 -1.03679167 0.3897917  0.23641667 0.42462500 -0.20070833 -0.28070833
6  -2.673542 1.7587708 -0.4745625 0.3505625 -0.66585417 0.4170000  0.16233333 0.22925000 -0.20750000  0.05270833
7  -3.243729 2.4683542 -0.1050625 0.3964583 -0.98029167 0.1623125  0.01958333 0.76229167 -0.03027083 -0.12239583
8  -4.051333 3.2339792 -0.1739792 0.3965833 -1.04602083 0.1951875  0.08666667 0.82077083  0.10445833  0.02122917
9  -3.876896 2.3450208 -0.3668333 0.3170417 -0.39450000 0.8033750  0.02504167 0.73614583 -0.23183333 -0.14810417
10 -4.506146 2.6885625 -0.2849167 0.4695625 -0.03879167 0.6388750  0.13916667 0.38756250 -0.11102083 -0.27335417
11 -2.990396 1.4638750 -0.5098125 0.3716458 -0.38039583 0.7250417 -0.08339583 0.50766667 -0.32750000 -0.22672917

Coefficients of linear discriminants:
              LD1        LD2        LD3         LD4         LD5        LD6        LD7         LD8         LD9
x.1  -0.904263484  1.0765784 -0.2138645  0.45014309 -0.01568403 -1.3197512  1.2563534 -0.55370976 -0.21893931
x.2   1.150256514  0.3481586  0.1219356  0.81974744  0.07879069 -1.2828586  0.8543263 -0.85921988 -0.46773378
x.3   0.539113617 -0.4775768 -1.3152720 -0.18174531  0.01428301 -1.0421541  1.3953387  0.05664837 -0.58464728
x.4   0.024636593 -0.6124198 -0.3333835  0.89249728 -1.19422174 -0.2840921  0.6079608 -1.18671349 -0.26235352
x.5  -0.007828209 -1.6078817  1.2374917  0.96925109 -0.17984293 -0.9791226  1.5595892 -0.12322124  0.38252584
x.6   0.708040285 -1.4820258 -0.8767861 -0.06187384  0.84899844 -1.9958303  0.2001485 -0.54892997 -0.82835032
x.7   0.843500553 -0.8675407 -1.4607005  1.10382558 -1.35657038 -1.0703157  0.5029121 -0.09927612  1.50030138
x.8   1.305220749 -0.9393954  0.9967222 -1.05415855 -0.85277476  0.1212728  0.6617860 -0.25729462  1.12742075
x.9   0.965050787 -0.5131795  0.6927861  0.60778087  0.04667900 -0.5487891  0.5244975  1.05461865 -0.04187258
x.10  0.352677857 -0.1517412  0.7092612 -0.22606164 -1.04293623 -1.7009401 -0.5629667  0.29898796  0.14972818
            LD10
x.1   0.01475366
x.2  -0.20534101
x.3  -0.52218214
x.4   0.22298959
x.5   0.05017199
x.6   0.13158053
x.7   0.36516484
x.8   0.30135489
x.9   1.31351481
x.10 -0.23691329

Proportion of trace:
   LD1    LD2    LD3    LD4    LD5    LD6    LD7    LD8    LD9   LD10 
0.5617 0.3518 0.0445 0.0191 0.0107 0.0083 0.0026 0.0011 0.0001 0.0001 
$\endgroup$
3
  • 1
    $\begingroup$ You had 10 X variables and 11 groups. Hence, you extracted min(10,11-1)=10 discriminants LD. It looks like "group means" are indeed themselves. Why not? "Coefficients" are the regressional weights to compute the LDs by the Xs. I can't tell, without having data, what is "proportion of trace", it may be related with the eigenvalues of the extraction. Please see my LDA of iris data. $\endgroup$
    – ttnphns
    Apr 1, 2014 at 9:49
  • $\begingroup$ Why do I have so many group means ? What are they group means of ? For example, for x.1, there are 11 group means ? Where do they come from ? $\endgroup$ Apr 1, 2014 at 16:32
  • 1
    $\begingroup$ Well, if you have 10 variables x 11 groups I guess there ought to be 10x11 means. $\endgroup$
    – ttnphns
    Apr 1, 2014 at 16:52

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.