Ray, thanks for your thoughts. This problem really calls for a careful analysis of what is implied by different study designs and ways of constructing samples and controls, and I'm glad you have raised this question. The concept of a control group is sort of problematic for my own study because it focuses on the question of how to arrive at a hypothesis, rather than on how to test one. It is a series of case studies rather than a sample study, which to me seems like a better fit for my current stage of investigation and the kind of knowledge I’m pursuing.
It's all very hard because we have virtually nothing that is astrologically solid to work with. We don't even have a rule in astrology that says people will gravitate towards each other - based on complimentary synastry. We actually don't have ANY clear-cut claims about anything in astrology. If we did, someone would be able to quote one of them ad verbatim.
I am convinced that people DO gravitate towards each other, based on complimentary synastry for good long-lasting relationships, but they also do it for less admirable reasons, such as 'gold digging', mental stimulation or getting a trophy partner or sex or simply finding someone who can tolerate them and help them live their chart and help them deal with their quirks, so I would expect that if we include all available couples in our surveys and pretend we are dealing with romantic relationships, we'll get an awful lot of background noise in our research graphs when we examine
This wouldn't matter a lot if astrological 'romantic attractors' worked at a fairly high percentage above the expected rate, but I've found that they don't - well not the 11,000 of them that I routinely use anyway. Ten percent above the expected rate is about the best I expect to ever see after processing so much data. So unless we can find some synastry factors that work at a fairly high rate, we have little hope of being able to start filtering our data in obvious ways after finding interesting peaks in our graphs.
Of course we can still do that - find an interesting peak in a graph, but one which is completely insignificant from a stats point of view, and then combine it with another equally high but also insignificant peak, and give ourselves a neat looking 'multi-factor'. Yes, it's ok to try that - just in case it might work, but it never has yet. I suppose in a case like that, the first factor may have been real, although not yet statistically significant because of the amount of data being used, but if we combine it with another factor which is only rating highly just by chance, then we don't find anything, but if we could somehow use all the permutations for that second factor, and one of them was real, we'd be able to move forward.
To put that in very simple terms - if we found that MER con JUP women married more men than expected with SUN in SAG or VEN in GEM, we may find nothing significant if we compared it with a control group, but if we tested for SUN
in SAG plus VEN in all the signs
, we might have success with (say) SUN in SAG or VEN in ARI, even though Venus in Aries didn't score particularly well in our set of data. As far as I know, no one is doing astrological research in this way yet.
Another thing that is worth considering, whether it's for your type of searching or mine - and that's the 'opportunity factor'. It's something that I have often wondered about, but I still cannot see it clearly enough. Lot's of factors are not present in transit events or synastry because it is impossible to get them in a particular era, or in the case of
partner-synastry, impossible to get within several years of the person's birth date, so I've been wondering how we can (in some surveys) isolate data that cannot produce 'hits' or can only produce a nominal number of hits. It doesn't seem quite right to me to be using data for people who CAN have certain factors in their synastry and those who cannot.
It sounds like your system of generating control charts is already handling that (apart from using a block of years, rather than a tapered-years' block, but I wonder if it might help a bit to also generate the controls with an allowance for the geographic coordinate boundaries
in which partners are usually found. For example if a woman was born in London, she would have a greater chance of marrying someone who was born reasonably close to her birth place or alternatively, her location at the time of meeting, so the control data should probably reflect that - even if some women who were born in London marry a foreigner in London.
What I'm driving at here is that the distribution of rising signs can vary so drastically in different parts of the world, that it can for example, reduce the chances of a woman in London with VEN in ARI marrying a man with Aries rising, because there are very few there. If she went to Australia she would find the place crawling with 'em and possibly increase her chances of having some very nice synastry
So I'm looking at individual cases and comparing one case with another, and this turns out to be a valuable exercise for clarifying some issues. In this scenario, there is no curve to match, only an individual partner. Even so, it is useful to apply a statistical model to a series of cases in order to see what stands out in a particular case. For instance, I can use Monte Carlo simulation to determine the probability of finding certain planets in aspect between a pair of charts. The question of what this means -- that is, in relation to what distribution of positions, or what time period -- is pretty open to various interpretations.
This case study approach begins with an astrological mindset, and so it highlights what is unusual about a given pair of charts. This is in contrast to the controlled sample type of study where the hypothesis states that some predictable or consistent relationship will show itself when looking across many charts. So it’s not easy to say how these approaches can be related to each other, but that’s the whole point of doing case studies this way. At least for now, I’m more concerned with how to use statistical methods in the context of astrological practice than with doing the hypothesis-driven type of study.
I think this is where other astrologers can help. They have had the opportunity to reflect on their own or their clients' synastry and often have a good idea about what brought them together. For example, in my parents' case, my (then) 30 year old father reluctantly went on a blind date and saw my (then) 18 year old mother standing at the top of an outside wooden staircase as she was about to descend. He told me many years later that he knew immediately that she was going to be his wife. She had her Venus con his Asc and Sun opp Vertex. There was probably more than that, but I never noticed. They split up in 1947 just after i was born in 1946, but there was never a question of divorce.
One more thing. it seems to me that relationship attractions are based on best fits - whether it's a dominant single factor, or a dominant factor plus other factors that get carried along with it in the sky by default, so if some people don't circulate very much amongst [realistic] potential partners, then they are reducing the chances of encountering someone with some of the best fits for their charts, and yet in research we assume they all had an equal chance.