4. Samples of Am type stars

We analyze now the stars classified as Am. Let us recall that Am stars are objects for which one quotes generally three spectral types- one corresponding to the strength of the Ca II lines, one to the strength of the hydrogen lines and a third one to the strength of the metallic lines. This precept has been followed by GG and AbM, but not by CJ, who called a star simply Am, without assigning the three types. CJ classify however a number of stars as Am:, i.e. marginal Am. GG follow a similar practice in introducing the proto-Am stars, stars which are metallic lined, but to a lesser degree. AbM do not quote marginal or proto-Am stars, although from the publication it is easy to see that such specimens do exist - like for instance an object classified B 9.5-A0-A1. Since these refinements are difficult to disentangle in a statistics based upon small samples, we have decided to apply the uniform procedure to consider Am stars everything which is called Am, Am:, marginal or proto-Am by the different authors.

For the theoretical prevision, the case is simpler than the preceding cases, since an object is either Am or it is not Am. Assuming that the classifier A has , the proportion of the coincidences should be 80% for , 70% for and 57% for . Here E(A) and E(B) design in a wide sense the accuracy of classification. To assume means simply that both A and B are equally good classifiers. means that A is better than B and so on.

We pass next to the real samples. From Table 1 (click here) it can be seen that GG had 50 Am stars in the sample, whereas AbM had only 38 and CJ, 24. When analyzed in detail, GG finds more early type Am's than AbM. The detailed comparison is given in the first part of Table 10 (click here).

 

GG: Am not Am

AbM: Am 57% 14%

not Am 28% Total number 70 objects 2) AbM and CJ

AbM: Am not Am

CJ: Am 62% 13%

not Am 13% Total number 68

 

A similar comparison can be made between AbM and CJ. The results are given in the second part of Table 10 (click here).

In view of the small number of objects involved, we can only state that about of the classifications coincide (close to the 70% prediction) and thus that the three classifiers have about the same precision. The difference of 14 and 28% of the first table of non-coincidences is at the limit of the statistical fluctuation, so that it cannot be used to decide which of the classifiers is better.

In what precedes we have neglected certain facts which also have an incidence upon the results- for instance that the MK standards used for classification are not exactly the same in all three samples, a fact which may lead to small differences. Such small differences are however difficult to pin point with (small) samples of a few hundred stars. Another fact is that the different plate dispersions used in the surveys could also have an influence upon the results. However since the standards used should be affected in a similar way than the objects being classified, it is hard to believe that this could introduce large differences.

Rotation is another factor which may introduce difficulties. GG and AbM have paid special attention to rotation. CJ did not pay special attention to rotation because the smaller dispersion used -125 A/mm- makes the classification less sensible to rotation, except for stars having larger than 200 km/s, which are rather infrequent in the range of A type stars.

On the other hand, it is gratifying to see that the results for both dwarfs and Am stars show that there are no big differences between classifiers, or in other words, that when the prescriptions of the MK system are followed,

a similar accuracy is reached, regardless of the classifier. Although this contradicts what has been written sometimes about the existence of more precise clasifications (i.e. more precise than others), our conclusion is entirely reasonable within the framework of the MK classification.