|
|
Dear SPSS users
I have recently been experimenting with the variance components option under the SPSS GLM. As I understand it, this facility is of relevance to random factors when trying to estimate to what extent these factors and interactions involving these factors contribute to overall variance in the data. I would appreciate clarification on the issues listed below.
1) Is it the case that the reason that the variance components option is restricted to random factors is that the variance components are to be understood as estimates of population variance and therefore the factor levels require to be understood as been selected from a population of levels, thus excluding fixed factors from having variance components?
2) Having successfully used the variance components facility to obtain variance components for a 3-factor fully crossed design in which all factors were random, I tried to obtain variance components for the same model but with only two of the factors random and the other fixed. Whichever approach I took (requesting a full factorial model or a customized model), the message I received in response was that no variance components had been generated because the residual variance was zero. I couldn't make sense of this message at all. In case this is of relevance, I should mention that I included the 3-way interaction in both the successful case in which all factors were considered random and the unsucessful cases in which two out of the three factors were considered random.
I would value some clear explanations.
Thank you in advance for your kind interest.
Best wishes
Margaret
---------------------------------
Inbox full of spam? Get leading spam protection and 1GB storage with All New Yahoo! Mail.
|
Locked
