Jump to content

stringer

Members
  • Content Count

    12
  • Joined

  • Last visited

About stringer

  • Rank
    Salisbury Hill

Recent Profile Visitors

284 profile views
  1. I can't remember now. It seemed a good idea at the time, when I was 14 years old! Chris
  2. I suppose some of his exploits were a bit weird. My childhood was a bit odd too, in its own way, and I could recognise a few traits in common. I had only two friends during my childhood and teen years, and almost no "normal" social interactions. I spent almost all my time when I wasn't at school shut away in my bedroom, which I had turned into a sort of laboratory, messing around with chemicals and electronics. There must have been blobs of spilled mercury all over the place, and quite a few other hazardous chemicals lying around. After an accident once when a pipe burst while I was making chlorine for an experiment (which, perhaps fortunately, terminated my plans for producing some highly unstable nitrogen trichloride!), I acquired a cough which still resurfaces sometimes to this day, decades later. These days, the house would probably be condemned until it could be extensively decontaminated! One of the things John Elder Robison wrote about that resonated very much for me was his feeling of being a "fraud" who would one day be "found out." I've had a reasonably successful career in scientific research, but still I have the feeling that I don't behave like a real "grown up," and that some day I will be found out... Chris
  3. That is a sad story, and the AS traits certainly come across in the account. It is interesting also to see how the style of newspaper stories has changed over the years. These days you probably wouldn't expect to find the home addresses of all the people involved in the incident reported in the paper. Do you know when it happened? The photograph of the film actress Susan Beaumont on the same page probably dates it to the late 1950s or so, I would guess. She was only active in films from 1955 to 1959, according to wikipedia. It must have been quite a discovery for you, when you found that cutting. Chris
  4. The 15 in 1000 figure is for persons of unspecified gender; it corresponds to the CDC estimate of 1 in 68 overall. Their estimate for boys (1/42) and girls (1/189) would then correspond to about 24 per 1000 for males, and about 5 per 1000 for females. Chris
  5. This is an interesting question, and I've not found any very clear answers in the searches that I've carried out. I would suppose that one way to estimate the fraction of people with ASD who are what one might call "low functioning" would be to compare the figures for the prevalence of autism in the past with the figures for the prevalence of ASD now. Thus, in the past it might well have been that almost exclusively those who were what would now be called "low-functioning" would have been included in the tally. According to figures I was able to find, autism was once thought to have a prevalence of somewhere between 1 in 10000 and 1 in 1000. Now, by contrast, ASD is thought to have a prevalence of about 15 in 1000. I suppose the majority of those extra cases in the recent studies would therefore be at the high-functioning end of the spectrum; that is to say, people who would, in the past, have slipped under the radar and would just have been thought of as eccentric or a a bit weird. I'm sure that is an oversimplification, but maybe not too far off the mark. I am assuming, in what I am saying, that there has been no actual secular change in the incidence of autism over the years, and that the changing statistics are entirely due to changing diagnostic standards and measurement. By the way, on the subject of books, I forgot to mention that Neurotribes, by Steve Silberman, is, I think well worth reading. He has some very interesting historical details about the roles that Leo Kanner and Hans Asperger played in the early research into autism. Chris
  6. I don't have much that is concrete to add. I have enjoyed reading various books in this general area, which I could recommend. John Elder Robison has written several. One of his, called "Look Me in the Eye," is about his own experiences growing up as an undiagnosed aspie. I found it quite gripping, because it felt in many ways as if I were reading my own autobiography. More recently, he has written an account, "Switched On," where he describes his experiences undergoing an experimental treatment called Transcranial Magnetic Stimulation (TMS), where areas of the brain are targeted with intense magnetic fields. I'm only part way through the book at the moment, but it seems to be quite interesting cutting-edge research that he participated in. A little like re-programming the brain, though at this stage with a bit of a blunderbuss approach. For other general reading on Asperger's, Tony Attwood has some interesting writings, which I found quite informative. Chris
  7. Tony, CDC is the Centers for Disease Control. They are the leading authority in the USA, I suppose, on all things related to diseases and conditions, prevention, data on prevalence, etc., etc. (Although British by origin, I've lived mostly in the US for many years now.) By the way, if the 1/49 you were alluding to is related to a number I mentioned earlier, that was a mistake on my part; I had misremembered the CDC figure, and I would have said 1/42 if I had rechecked their figures at the time I may have said 49. Yes, that little hump over to the left on the solid curve does look curious. But with a total ASD sample size of only 58, it's really down in the noise, I suppose; just two or three people, perhaps. It wouldn't be hard, indeed, for someone to distort their AQ scores if they wanted to. Who knows? It would have been very interesting if in that 2001 study they had gone back and tested (with a full diagnostic test) the two or three individuals in the "controls" who were skewing things at the high-scoring end, to find out whether they in fact had ASD. About whether there is some upper limit to the AQ scores for those who qualify as HFA, I'd like to think the answer is no; my AQ score is 44, and I think I manage OK! I think HFA can span the whole range right up to the highest score, probably. I don't know, though, what proportion of the people with ASD would count as having HFA. My guess, though, would be that the vast majority of the 1/42 fraction of the male population that would now be classified as having ASD would fall into the category of HFA. I say that simply on the basis that the latest estimate of 1/42 is so very much larger than the fraction that would have been recognised in the not-too-distant past. I would assume, therefore, that the bulk of the extra people in the new classification would have passed for just being "a bit weird" or eccentric, in earlier days. So I think you have plenty of chance to meet some in the street! Chris
  8. That is a very good point that you make; the "controls" in the reported study were not verified not to include individuals with ASD. I had not read the paper carefully enough. I suppose the presumption would have to be that the controls would include people with ASD, perhaps with the standard prevalence in the population at large. This could have a very large impact on the nature of the high-scoring tail in the NT distribution function. One would need to subtract out the effect of the ASD individuals from the "control" distribution, in order to get a true NT distribution. Much of the present high-scoring tail in the control distribution may go away in the true NT distribution. In turn, this will greatly increase the probability that a score s in the range from around 30 and upwards corresponds to a person having ASD. In principle, given enough data from a very large number of "controls," one should be able to quantify this correction, and come up with a true NT probability distribution. But the numbers of individuals in the control group in that 2001 study was much too small to make that possible. In that set of 174 controls in their study, there might on average be expected to be two or three with ASD, but the uncertainties in such small numbers are too big for this to be reliable. But it could well be consistent with removing the tail on the right-hand side of the dashed curve, giving something looking much closer to a bell curve. Chris
  9. Yes, one could say that it is a little misleading displaying those two plots on the same diagram, in the sense that a random person is much more likely to be on the dotted NT distribution curve rather than the solid ASD distribution curve. Of course there is nothing actually wrong with plotting the two probability distributions on the same diagram. It's just that one is then tempted to make a direct comparison of the areas under the two in the way you were describing, whereas in fact the area under the NT curve should be scaled up by the large factor (like 42) relative to the area under the ASD curve. So I would question your final statement in your item 1 up above. Without the additional knowledge of the proportion of the population that has ASD, you wouldn't know how likely it was that there might be any ASD people among those random 20 people in the room. In my extreme example I discussed above, where we temporarily pretend that the occurrence of ASD was at the level of one in a million, the 20 people would almost for sure all be NT, and all that you could say is that you were probably among the top 5% of high-scoring NTs. In fact, with the ASD occurrence being more like 1 in 42 (for males), of course things are less black and white, and scores like 32 are in a rather murky region where details of the tails in the probability distributions matter a lot. I can imagine that a tricky point if one did want to determine the NT probability distribution more precisely, by using a much larger sample size, is that one would want to be sure that the high-scoring tail in the "NT distribution" really was due to genuinely high-scoring NTs, as opposed to people with undiagnosed ASD who were being mis-classified as NT when they were not. Bayesian statistics are all about refining one's estimate of probabilities, based on prior knowledge of related conditions. Chris
  10. Well, there is the additional issue, which is not addressed purely by looking at the two distributions in figure 1, of what proportion of the population has ASD. For this one needs further data. Suppose we take an absurdly extreme example just for the sake of illustrating the point: Suppose that it were the case that just one person in a million in the general population had ASD. That would then mean that a random person with an AQ score of 32 would be overwhelmingly more likely to be an NT at the high end of the NT scale rather than being a person with ASD at the low end of the ASD scale. Of course, my example was deliberately absurdly exaggerated, just to make the point that the knowledge of the proportion of the population that has ASD must be factored into the calculation. In actuality, if we take the latest CDC figures for the prevalence of ASD among children, it is 1 in 68 overall. It is also said to be about 4.5 times as common in boys as in girls, and hence they arrive at the statistics that 1 in 42 boys has ASD and 1 in 189 girls. Presumably, there is no reason to think that these figures would have been significantly different in the past, and so that would mean that in any age group about 1 in 42 males would have ASD, and 1 in 189 females. If one knew the probability distributions for AQ scores in the NT and the ASD populations with precision, it would then be a simple matter, with Bayesian statistics, to calculate the probability that a given AQ score s corresponded to a person having ASD. The problem in practice is that the probability distributions that can be found in that paper are rather approximate, being based on quite small sample sizes. Especially, that all-important "tail" at the right-hand side of the NT distribution in figure 1 is actually based on the AQ scores of a tiny number of individuals. But knowledge of the details of that tail are rather important for determining what proportion of people scoring, say, 32 in the AQ test are going to be high-scoring NTs in the relatively large NT population, versus lowish-scoring ASDs in the much smaller ASD population. It is possible that in my estimate of the probability distributions, which I was trying to determine from the plot in figure 1, I may have somewhat over-weighted the high end of the NT scale. I'll try to take another look in the next few days. Chris
  11. Yes, Arlene, I think indeed there is sometimes a tendency in the popular discussions to overestimate somewhat the significance of a given AQ score, for the reason I described. By contrast, I think the research articles by Simon Baron Cohen and others are rather measured and accurate in what they are saying. And I agree absolutely with you that statistical analyses of data from a screening tool like the AQ test cannot in any way be taken as a diagnosis. But a useful guide, and something perhaps to spur someone into further research and reading about ASD, and to see whether they have the other traits too. Perhaps also as a stimulus for pursuing the possibility of a formal diagnosis. Bearing in mind that the CDC estimates that about 1 child in 68 has ASD, and that this fraction has probably not changed over the years, that means there must be very many undiagnosed adults out there. So yes indeed, Tony, it is quite likely that we are rubbing shoulders frequently with other undiagnosed adults. For most, it is probably not realistic to hope that they will ever receive an evaluation and formal diagnosis. But for many people, like me and, by the sound of it you too, the realisation that we quite possibly do have some form of ASD can be quite liberating, I think. Finally, one sees a pattern to some of the difficulties one has had over the years, and some level of understanding of why some things like socialising and forming friendships and relationships can be quite challenging. Being of a scientific background myself, I find it quite interesting to try to extract as much information from the available data as possible. Chris
  12. Gigaday, that is a very interesting question you have asked. I thought about this issue too, and I do have a couple of observations. The question I asked myself is "What is the probability that a randomly selected member of the population who gains a score s on the AQ test has ASD (Autism Spectrum Disorder)?" To answer this, one needs to know the distribution functions for AQ scores among both the population of people with ASD and also for those without (NT, neurotypical). Given those two distributions, it is then a rather straightforward statistical problem to work out the probability that a random person (might be ASD, might be NT) who scores s is in fact ASD. This is purely statistical, of course, and not diagnostic for that particular individual. The original 2001 paper on the AQ test by Baron Cohen et al., is available at the Autism Research Centre website, http://docs.autismresearchcentre.com/papers/2001_BCetal_AQ.pdf (sorry, you probably need to delete after the pdf at the end of the http address) In this paper, one can get a rather rough and ready answer for the two probability distributions, in their figure 1. I took those plots, and tried to estimate distribution functions that essentially modelled those curves. Then using these to perform the statistical analysis, I came out with estimates for the probability that a person scoring s on the AQ test has ASD. The results are significantly different for males and females, because of the substantially different prevalences of ASD in males and females. I think one of the greatest uncertainties in the results I was getting stem from the fact they depend rather crucially on the details of the "tail" in the NT probability distribution at the high-scoring end. The point being that although the probability that an NT gains a high score is rather low, the probability that a random person is NT is high (about 49/50 for males, if one takes the latest CDC estimates for prevalence of ASD in the general population). So even a rather small tail in the high-end of the NT probability distribution can have a quite significant effect in lowering the eventual probability that a score of s for a random person implies that they have ASD. Anyway, after all the caveats about the inaccuracies of the the estimates, here is what I got. For the various AQ scores listed in the first row, my estimates for the percentage probability that that score corresponds to the person having ASD is given for males in the second row, and for females in the third row: AQ score: 28 30 32 34 36 38 40 42 44 46 Male: 2 5 11 20 33 46 59 68 75 80 Female: 1 1 3 6 10 17 26 34 43 50 Of course there have been many studies since 2001 in which groups of people have taken the AQ test. Quite a lot of these have been published in the peer-reviewed literature, and so in principle there should now be a lot more reasonably reliable data available, allowing one to make much better estimates for probability distributions for AQ scores of the NT and ASD populations. If I have some spare time in the coming months I may have a go at doing this. One of the reasons I tried making this analysis was for my own personal purposes. As an undiagnosed adult with a very high AQ score, I was curious to try to estimate the a priori probability that I have ASD, since this could influence whether I want to go through the effort and the expense of getting a professional diagnosis. The estimates I got are probably fairly rough and ready. But I haven't seen any other estimates anywhere that attempt to take into account, in a quantiitative way, the important effects of the NT distribution function in the way I have tried to do. Chris
×
×
  • Create New...