В наше время методологи считают, что устойчивость факторного решенияне зависит от соотношения количества переменных и испытуемых, важно толшько, чтобы выборка была как можно больше:
Many sources suggest that a certain minimum ratio of sample size to number of variables (e.g., 5 to 1, 10 to 1, etc.) is needed to ensure the stability of factor analytic results. These suggestions are entirely misguided. Having a larger number of variables in a factor analysis does not undermine the stability of factor analytic results, and neither does the ratio of sample size to number of variables influence the stability of results. Because the sampling error of a correlation coefficient depends on the sample size rather than on the number of other correlation coefficients being calculated, the factor analysis of a large variable set does not require an especially large sample size (see Guadagnoli & Velicer, 1988, for an empirical demonstration). (Handbook of research methods in personality psychology, 2007. Стр. 424, авторы Kibeom Lee и Michael C. Ashton)
Поддерживаю выкладку alexwin1961. Рекомендация, уже издавна известная, гласит, что с увеличением в ФА числа переменных максимальное приемлемое соотношение число пер-ных / число набл-й растет. Так, если пер-ных 20, N желательно никак не меньше 100, но если пер-ных 100, N вполне достаточно 300 (а не 500). И дело здесь не только в надежности r, а в том что для "хорошего" ФА нужно, чтобы признаки достаточно широко и плотно "покрывали" изучаемое поле свойств: чтобы частные парные корреляции были низки. Поэтому, в общем, чем больше пер-ных, тем лучше для ФА.
> тобы частные парные корреляции были низки добавлю: те частные корреляции, которые случайно отличаются от нуля. Увеличение размероа выборки не поможет, если частные связи имеют место в генеральной совокупности
Нет, я немного неясно выразился. Говоря о частных корреляциях, я уже ушел от вопроса размера выборки и имел в виду просто сказать, что обычно мы вообще не хотим сколько-нибудь высоких частных корреляций, когда полагаем использовать ФА.
Many sources suggest that a certain minimum
ratio of sample size to number of variables
(e.g., 5 to 1, 10 to 1, etc.) is needed to ensure
the stability of factor analytic results.
These suggestions are entirely misguided. Having
a larger number of variables in a factor
analysis does not undermine the stability of factor analytic results, and neither does the ratio
of sample size to number of variables
influence the stability of results. Because the
sampling error of a correlation coefficient depends
on the sample size rather than on the
number of other correlation coefficients being
calculated, the factor analysis of a large variable
set does not require an especially large
sample size (see Guadagnoli & Velicer, 1988,
for an empirical demonstration).
(Handbook of research methods in personality psychology, 2007. Стр. 424, авторы Kibeom Lee и Michael C. Ashton)
Reply
Reply
Reply
добавлю: те частные корреляции, которые случайно отличаются от нуля. Увеличение размероа выборки не поможет, если частные связи имеют место в
генеральной совокупности
Reply
Reply
Leave a comment