 |
 |
Dear Sun, What about cluster-level inference that I suggested to you? Was anything significant at the cluster level? In general, you can continue using SPSS to look at components which are known from the literature. SPM can help you discover new effects not necessarily limited to peaks (in principle there is no reason why interesting effects should be at the peaks and not all peaks actually have physiological meaning). However, the price you pay is that you have to be more strict and perhaps should include more subjects in your study than you would with a standard SPSS analysis. If you discover a new effect using SPM you can test it with SPSS in your NEXT study as long as you are not using the same data for SPM and for SPSS. That was the part I opposed. Best, Vladimir On Tue, Nov 2, 2010 at 2:38 AM, SUBSCRIBE SPM Sun Delin <[log in to unmask]> wrote: > Dear SPMers, > > Thank you very much for your excellent discussion on the statistical meaning of SPM. I'v learned a lot from your messages. I am not familiar with SPM' statistics, however, I feel that the statistics in SPM is much different from that in SPSS I used in my ERP data analysis. I have found several ERP components in my study significant by using SPSS, however, when I did the similar analysis by using SPM, some components did not show their significance and some were less significant (even without correction) than the SPSS results. I was surprised to find that my most significant result, which was consistent with a plenty of previous studies, could not survive after the FWE correction (p < 0.05). That's why I ask about the standard of significance used in SPM for ERP data. I am afraid that many previous studies of ERP were not so strict in view of SPM that they were full of false positiveˇ]though the findings were consistentˇ^, and many theories based on previous ERP studies should be tested by using more strict criteria again. On the other hand, I am also afraid that the ERP experts by using traditional statistical methods would not accept SPM's results for their complex statistics. Therefore, I suggest a way to use SPM as an "exploratory" way to detect the possible significant results and then show results by using SPSS (However, this idea was opposed by Vladimir intenselyˇA ^_^). I am hesitating how to use SPM on ERP data now. > > Sun Delin >