Which online Asperger’s tests

Post Reply New Topic

Ok... I'll admit, you got me there.

I'm aware of RDOS's neanderthal theory and while it's possible, I don't agree with it. There are parts of the RDOS test which particularly relate to that theory (particularly "hunting") and I don't think they add anything to the test. RDOS doesn't push those theories too hard. (or at least not that I've noticed).

By comparison, Baron-Cohen's male brain theories have been pushed very hard and I disagree with them more than RDOS. I'd also heard that he was backing away from that theory now - or perhaps that was just wishful thinking...

In any case, I wasn't aware that Wired had the whole test. If that's the case then it's a bit short for my liking.

I remember most of the RDOS questions from my own testing.

This is probably true (according to how AS currently is defined). The thing is that Aspie-quiz was not designed as a diagnostic instrument for AS. It is designed to identify Aspie traits that are not secondary psychological problems in adults. Because of this intention, later versions have lower validity for AS-diagnoses than some of the earlier. However, as I also wrote in the paper, there is a trend for lower validity in latter versions, a trend that cannot fully be explained with question selection. This trend quite likely exists because the increased popularity of Aspie-quiz means that more people in doubt about their diagnosis have done Aspie-quiz, and less with no doubt. I don't even think the issue of Aspie prevalence or validity for diagnoses is properly evaluated, as this is impossible to do with the non-random population selection that exists. Some guesses about the prevalence of "very likely Aspie" can be done by looking at near-random forums, but those have big uncertainties (a guess is 5-15%).

So, which one is most reliable depends on what people want to know. Aspie-quiz is far better at identifying autistic traits in adults than the AQ-test, and it also give non-gender biased results. Aspie-quiz also uses natural trait linkage as a method for scoring, and not reasoning-based decsions about which traits should be in DSM and which should not. As a research tool for autistic traits Aspie-quiz is also far better, especially since it is much broader. The AQ-test is better for identifying AS as it is currently defined in the DSM, simply because it was constructed from DSM.

Probably need to comment on this also. I've read this study, and I've compared the scores in his control-group with scores in the average Aspie-quiz control-group. Cohen's control-group is not even close in scores to a typical control-group in Aspie-quiz, meaning that they comprise of extremely neurotypical people, like for instance students in social sciences, which is common in these studies. When he selects his control-group in this way, it is natural that he has seemingly high discriminant power. People in the control-group in Aspie-quiz have much higher scores, even if they are selected based on forums unrelated to ASCs.

OTOH, if anybody is interested, it is perfectly possible to check differences in discriminant power between Aspie-quiz and the AQ test using the results from version R4 and F1 of Aspie-quiz. Here, the same people took both tests, which is the only valid way to compare. And the sample is much larger than Cohen's (between 1,500 and 2,000).

No, the AQ test does NOT have better discriminitive power for AS. I've now included the AQ test in the referral reports for version R4 and F1 and they can be studied here:
http://www.rdos.net/eng/aspeval/refr4.htm
http://www.rdos.net/eng/aspeval/reff1.htm

It turns out that I was perfectly right about the suspicion that many of the diagnosed people are in doubt about their diagnosis. In F1, only 61% in the diagnosed group score "very likely Aspie", but only 59% score 32 or more on the AQ test. In R4, which was before Aspie-quiz was well-known, 63% score "very likely Aspie" while 81% score 32 or more on the AQ test. Also, 16% in the neurotypical control group score in the AS range on the AQ test. It is also evident that females generally score lower on the AQ test in comparison with Aspie-quiz.

Also, looking at the whole group in F1, 43.5% (42% for males, 45% for females) score in the AS range on the AQ test, but only 34% (31% for males and 37% for females) score very likely Aspie. This actually means that Aspie-quiz has better discriminative power

A minor methodological flaw in the last results. I compared the whole group against people that did the AQ-test, which is not entirely appropriate

Anyway, I changed logic and now the results are for identical populations. The differences are minor, but the Aspie-quiz scores have gone up. In F1, diagnosed AS now have 70% "very likely Aspie", while the AQ test still has 59% in the AS range. The neurotypical control-group has 19% "very likely Aspie" and 16% in the AS range on the AQ test. These results seems to show that slighty more people get "very likely Aspie" than AS range in the AQ test.

I'd say Aspie-quiz still has better discriminative power

This is incorrect. The control group in the AQ study that gives the result of only 2% scoring above 32 consists of a random sample of 174 adults.




When comparing results, only those diagnosed AS/HFA must be used; the AQ studies exclude those with an IQ below the normal range.

The figures from your F1 (of 70% and 59%) appear to be obtained from those diagnosed with the entire range of ASDs, not just AS. If your Aspie control group is used, the figures become 71% "very likely Aspie" and 69% for the AQ test. They are ~ equal.

That the result for the AQ is 69% as opposed to around the 80% value in the literature does suggest some sampling biases, like you mentioned. The NT control group strongly suggests this, with 16% scoring above cut-off on the AQ, as opposed to 2%. However, comparing the NT scores, having 19% score above cut-off on the Aspie Quiz again indicates that this version does not have better discriminant power than the AQ.

Looking at R4, your Aspie control group gives 66% "very likely Aspie" and 79% for the AQ test (~ agreeing with the literature). Therefore, the AQ had better discriminant power than this version. I don't see an NT control group.


Just to check; is your cut-off for "very likely Aspie" always kept at the AQ equivalent? Do you take the AQ cut-off to be 32? What I wrote above depends on this assumption.

OK. I still see big problems with recruiting a random sample of adults. The only group that resembles such low scores I've ever seen is the psychology-students at NTNU. Not only do you need to select a random group, you also must get back the survey from alomst everybody. If you for instance only get back half of the surveys, interest will play a major role. This is exactly how recruitement from random forums online work. They give much higher scores than a random sample.



They are here also. The autism group is separate, and it also has a magnitude of lower participation.



It is AS/HFA/PDD. At least that is what is asked for.

Besides, in R4 the autism group is separate, but scores are very similar (it cannot be ruled out that they are identical).

The Aspie control group is autism-community referral sites. It is not valid here as most are undiagnosed.



Yes. It suggests that selection among the diagnosed sample is not random. It actually suggests that people in doubt about their diagnosis are overrepresented, and that some of these people are misdiagnosed.



Absolutely. It is people that have suspicion about having AS that primarily do Aspie-quiz, which gives a skewed participation rate and much higher scores in control groups.

What does indicate better discriminative power is that in the diagnosed sample Aspie-quiz have consideranbly higher scores, while scores in the control-group are almost identical. 11% more get "very likely Aspie" in the diagnosed sample, while only 3% do in the control-group. In R4, the AQ test give 7-8% more "AS indications" in the whole sample, while in the diagnosed sample it is only 6% more. This does suggest slightly higher discriminative power of Aspie-quiz.



Nobody in the NT control group answered to the AQ test, so we need to use the whole sample instead. As I wrote above, in R4, the AQ scores are higher than the Aspie-quiz scores.



No, the cutoff was selected years ago to be Aspie score - NT score >= 35. It is a pure coincidence that this roughly corresponds to the AQ cutoff.

How do you guarantee that when selecting whether they are AS/HFA/PDD that it excludes those with IQ below normal range? Even if there happened to be a low percentage of below normal IQ responders in this group, it's best to minimise contamination between the groups.



Perhaps a separate group for PDD should have been made to prevent test-takers with PDD-NOS plus below normal IQ, or milder ASD, affecting any comparisons of discriminant power for AS/HFA.



That's something to be looked into with higher numbers. However, if this group consists of LFA as well as HFA responders, that would be a mixed IQ group and (since IQ is not controlled for) could not be compared to the results of any significantly higher IQ group. If they were just LFA, they still cannot be compared to a different IQ group. Was there any intention here to separate HFA from AS?



Why was it was named Aspie control group?



Your results might indicate higher discriminant power in this version, but the PDD group here might be too different from the AS/HFA group the AQ was designed and tested for. Is there any way to determine this; I recall you found a g factor somewhere? It seems more likely the sample is biased by milder ASD cases, though.



Is a 7-8% difference as opposed to a 6% difference significant? You seem to be saying that, in R4, the AQ gives higher indication of AS in both samples (one of which contains the diagnosed group) than the Aspie Quiz does, with a pretty consistent difference, yet has lower discriminant power, which wouldn't make sense.


Some inquiries: In R4, it appears you are referring to the whole sample as the Male (N=298) plus Female (N=350) groups, which totals 648. This then appears to consist of the following groups:

Male + Female Diagnosed AS/HFA/PDD (N=385)
Autism (N=12)
ADD/ADHD (N=60)
AS control (N=191)

Is this assumption correct? If so, why were self-diagnosed excluded?

Also, the Diagnosed AS/HFA/PDD group consisting of 68 near the top of the table appears to be unrelated to the Male + Female AS/HFA/PDD groups (total=385). Was this group some sort of subset within them or was it completely separate and not taken into account in the group you referred to as the whole sample?

It would surprise me a lot if IQ has any relevance. Aspie-quiz is not loaded on IQ to any great extent, and thus IQ should be irrelevant. There is a third factor in Aspie-quiz that I sometimes refer to as the "g-factor", but it can only explain about 1% of the variance. The Aspie factor explains about 2/3s of the variance.



Given that the ADD/ADHD and social phobia group shows almost as high correlation with scores as AS/HFA/PDD, I don't see how breaking up ASCs even more would make any sense. You might remember that in the neurodiversity versions I tried to create diagnostic factors for a set of DSM diagnoses, but the end result was always that if you got one you often got all of the others as well. That's why I gave up on that. It didn't work. There were no natural borders that could be exploited, and the only way to get some success would be to use official criteria for the proposed dysfunctions, which I didn't want to do.



Since the neurotypical control group is a set of hand-picked non-autism related forums, it made sense to hand-pick a set of autism related forums and calling that for the "Aspie control group".



Possible. Anyway, Aspie-quiz never have targeted ASCs specifically, and much less individual ASC diagnosis. Instead, relevant traits from related diagnoses have been tried and some have been incorporated. If the AQ-test does specifically target AS, it should have better discriminative power, which seems quite unlikely to be the case. Such a failure of the AQ test could be related to many things, but I suspect the primary reason is that AS-diagnoses mostly are related to natural factors of human variation, which Aspie-quiz measures, and because of this Aspie-quiz can actually do a better job without even looking at the DSM.



I'm not sure if this difference is significant. Could very well be that it isn't. But R4 was an experimental version. F1 is the first final version, which is the one which should give the best results.



No, the whole sample is everybody that answered the AQ-test. The Male and female AS/HFA/PDD group consists of both self-diagnosed and professionally diagnosed individuals grouped by gender (it also consists of people diagnosed / self-diagnosed with autism). Only using diagnosed AS/HFA/PDD and grouping this on gender gives a too small sample. The ADD/ADHD group only consists of diagnosed individuals (which should be indicated by the name, but I think this was changed as I wrote the paper).

Sorry to disapoint you if you were thinking you were Aspergic as a reason for why you are as you are. But Childeren with AS grow up into adults with AS there is no cure or fix. You just learn to deal with it, and the test you mentioned is very biased toward aspergers, I know I took it and it said I was higher spectrum than my actual diagnosis from the psychiatrist, so if it says no you dont have it you probably don't. Sugest if you feel out of place or want to know if you were affected when you were younger and have maybe just dealt with it well in adulthood, try get your doctor to test for similar disorders. But I don't think it's specifically AS you have.

Arguably the new 'Gold Standard' for Asperger testing is the Cambridge AAA test which incorporates several questionnaires and a whole range of generalised questions. Available to view from the Cambridge Autism Research Centre website (easily found online from those words). It's Simon Baron-Cohen's centre.

MeowMeow
That was a quick diagnosis
What else similar can it be then?
I don’t believe in some superficial 50 question online test.

I have been reading carefully what rdos and outlier have been writing, don’t understand much of it.
If 80% of diagnosed scored over 32 points. Does it certainly mean 80% of diagnosed women and 80% of men or is it 80% of all the diagnosed. As 80% of diagnosed are men.

Featherways
Is it possible that you give a link to the test

I’m gonna get tested anyway. Hope it doesn’t take 8 years.

But I have taken it as part of my identity now that I have Asperger’s. Even if I don’t have, I can identify myself with it because it has given me all the big answers what I never understood about myself.
Also someone in my family has it, so I might understand him better.

I think I found Cambridge AAA test

http://www.autismresearchcentre.com/tests/default.asp
it’s a good link.

but it’s not online test, you have to download tests and google for answers.

If you want to specifically compare the discriminant power of each test in correctly identifying AS/HFA, which I thought was what we were comparing (because the AQ and its cut-offs were designed for that), the diagnosed sample must consist of only diagnosed AS/HFA.



If the Aspie Quiz identifies autistic traits in subclinical and milder PDD cases and classifies more of them "very likely Aspie" than the AQ, then it wouldn't be surprising there is a large overlap with SAD, ADD/ADHD; and larger than if it was just designed to classify those reaching the threshold for an AS/HFA diagnosis. There will never be clear borders, but there can be significant differences between groups consisting of AS/HFA, milder PDD, and things like social phobia. Even one AQ study (the Dutch one) found this to be the case.

Anyway, again, I only mentioned reducing contamination due to PDD-NOS and low IQ cases above because I thought we were discussing the relative power of the tests to correctly identify ASA/HFA (a diagnosis of which requires more/severer traits/symptoms to reach threshold than milder PDD).



Maybe put a note somewhere in the table about how the groups were selected.



I don't see why you think the AQ is unlikely to have better discriminitive power and refer to its failure: Again, if we are discussing a certain level of traits--that required to meet the threshold for an AS/HFA diagnosis--then the AQ was designed for that group and the cut-offs chosen according to that population. If we are discussing which test can better discriminate subthreshold and milder cases, then Aspie Quiz is probably the test to use.

Without having clean data to analyse, the relative discriminant power of the tests in identifying AS/HFA cannot be determined.

Have you performed a ROC analysis on Aspie Quiz? If you wish to compare the abilities of the 2 tests to correctly identify AS/HFA (excluding PDD-NOS, self-diagnosed etc.), then, assuming the samples could be cleaned up to have a group containing only AS/HFA and this analysis performed on the new data, the AUCs could be calculated.

"Natural factors of human variation" seems to be just be another way to refer to the spectrum of autistic (and other neurodiverse) traits. The professionals acknowledge there is a spectrum (or, more accurately, multidimensional continuum) blending into normality, but have to make artificial cut-offs for people experiencing clinically significant levels/difficulties. Aspie Quiz would of course do a better job at identifying these "natural factors of human variation" if it is designed to allow milder and subthreshold "caseness" to be identified, as well as traits not very highly correlated with autism.

Yup. It's meant for diagnostic professionals to work through and input the answers they take from people during the process, then the computer comes up with its rating and they compare theirs as a Professional and come up with a conclusion. Because it's so broad-ranging, I think it's pretty robust.

I don't find these tests very well thought out. For instance, the EQ-test, which is supposed to measure empathy, in realilty have of core of questions that in autistics will get positive/negative answers not because of non-existent empathy, but because of problems to read emotions of others, especially IRL. These are mixed-up with real questions about empathy (which I doubt have any relation to ASCs whatsover), and totally unrelated questions. Baron-Cohen should first have checked the relations between questions, and if they are relevant at all for ASCs before putting this into a "Gold Standard". I also know that several questions in the AQ test have zero relevance for ASCs.