 |
 |
Dear Nan, It is true that the automatic HP estimation can be very conservative sometimes, and it sounds like that is the case for you. Normally I'd expect that it wouldn't matter so much given your longish blocks, but it is certainly a possibility that getting the right value can make a difference. I think you are modelling things correctly, by splitting into separate EVs, and I would not change that. What I would do to find out a good HP cutoff value is to start with a large value (e.g. 500s) and then progressively decrease that value and look at the effect that it has on your EVs (just hit the "View design" button after changing the cutoff value). When you start to notice a change in the EVs is the point at which you should set the cutoff. Basically you want to avoid the HP filter affecting your EVs, and in particular you want to avoid it removing power from your design, which is reflected by altered EVs. If you get down to 90s and you still don't see a change, then stop there, as going to values less than 90s can cause problems for the autocorrelation estimation. If you follow the above method you should be find an appropriate cutoff value. All the best, Mark On 14 Mar 2014, at 11:58, Nan Wise <[log in to unmask]<mailto:[log in to unmask]>> wrote: Dear experts: MY DEFENSE IS SCHEDULED IN 8 WEEKS ! I am having a hard time selecting an appropriate HP filter for a 20 minute experiment that examines the difference between imagined stimulation of the nipple and clitoris versus actual physical stimulation of those regions. HERE'S MY DESIGN For my 22 minute imagine-map experiment the design is as follows: 60 sec rest imagine nipple model (that means they are imagining making the hand movements they would make if touching their fingers but not touching their nipples-- a control for the nipple imagine) 30s imagine nipple stim --i(that means they are imagining touching nipples--30 s alternating for a total of 5 mins then imagine clitoris model 30s imagine clitoris stim 30 s alternating for a total of 5 mins then nipple model (actually touching fingers together but not touching nipple) 30 s nipple touch 30 alternating for a total of 5 mins then clitoris model (touching fingers together but not touching clitoris) 30 s clitoris touch 30 alternating for 5 mins then there is a 30 s imagine speculum 60 s imagine dildo 30 s speculum imagery one time each so that the total time for this is 2 minutes then 60s rest at the end. As per the recommendation of an advisor, I am setting up a separate EV for each of the model and stimulate conditions whether it is imagery or actual physical stimulation. If I use the estimate HP setting, it estimates a very long HP --well beyond the length of the study in sec. Say the study is 749 TRS which makes it 1498 s--the hp estimate wants me to set it at 13,482 s. Does that make sense? A colleague recommends a HP setting of the same amount of secs as the study so in that case my HP would be 1498. An advisor suggests using the default HP = 100. And in the past I have used 60 s when I have thought of the design as a block type with 30 s on and off and haven't set up EVs for the "model" conditions. I am looking at data run in these different ways. I get quite different results as you can imagine with the different filters I use. In general, the higher the hp, the more activations, including areas around the skull, etc, that may in fact be noise. I am getting dazed and confused. I need to tie up this analysis in the next few weeks as I have my defense schedule for early May. Is there a good way to look at how the data is fitting the model in terms of the time series plots? Would that be a good barometer? Another question: Would it be best not to set up EVs for the model condition and basically treat it as a true block design--rArA? I AM VERY MUCH APPRECIATIVE OF ANY HELP YOU CAN OFFER Nan Wise Rutgers-Newark Nan Wise, Ph.D. candidate Cognitive Neuroscience, Rutgers-Psychology Certified Sex Therapist, AASECT Fellow, The American Psychotherapy Association Fellow, The National Board for Clinical Hypnotherapists Board Certified Diplomate, The American Board of Examiners in Social Work