Reading is Fundamental,

I am pretty sure that you are aware there are better solutions than that.

If you want to compare two different questionaires, you give the tests to two representative samples of the population you plan on testing, and then do a paired t test.

If for some reason being able to compare your rates to previous years accurately isn't even worth the effort to do that, then you add the new question that specifically asks about autism , to the old questionaire, so you are effectively mentioning autism twice. Not perfect, but extremely cheap.

If for some reason you want to make sure you have a semi-plausible out when rates that you wanted to stay stable show an increase, then you do neither of those things.

And after what you describe as a massive undertaking, we are still left with the question as to why the data describing boys rates of autism by year was somehow not put into the data tables.

The trouble is, that as John Stone points out, the people really exist, and statisticians can't wave a magic wand the way they can with numbers, to make actual people go away..

An inaccurate study, or one that can't be used to compare rates to previous years, isn't just marking time while people save face and quietly change things to prevent further harm. Instead, people look at studies, take them at face value, don't read the fine print, and then make policy decisions ( like adding more vaccines during pregnancy) and can't understand why the rates of autism keep increasing because a scientist they trusted told them everything is fine.

Numbers on paper translate into real peoples lives.

If you felt that a health condition was important to track, and you wanted to change the words on the questionaire, *( which is what they did, asked a specific question about autism diagnosis instead of including it in a question that listed other disorders too) wouldn't you run a comparison with both questionaires to indicate whether changing the wording on the questionaire actually caused any significant difference in responses or not?

You mean do the NHIS twice? No, I wouldn't. It's a massive undertaking. The technical aspects are set out in excruciating detail, e.g., here for 2014 (see p. 52). The variable CCONDL06 has been replaced by the variable AUTISM. Do you have any specific complaint about the wording of the item?

Hi Greg,

Thanks for your kind words ( and I am not a statistician, so please anyone else chime in!). Yes, adding the numbers up it does seem like there is some overlap between the three categories. I also suspect that environmental and vaccine injury harm is going to show up in multiple ways, not all of them related to autism.
My personal belief in the good faith/integrity of these studies went out the window with the confession by Dr Thompson that data was dumped in trash cans to hide significance.Hard to get past that.

What actually really strikes me about the data though, is what is not included.
A couple of questions for RIF.
If you felt that a health condition was important to track, and you wanted to change the words on the questionaire, *( which is what they did, asked a specific question about autism diagnosis instead of including it in a question that listed other disorders too) wouldn't you run a comparison with both questionaires to indicate whether changing the wording on the questionaire actually caused any significant difference in responses or not?

Why would they not do that? It's pretty basic science.

Now let us imagine for a moment that a scientist was seriously interested in finding out whether an increase was significant or not. They would want to look at what happened to the sub population that showed the most increase, wouldn't they?
(Boys, specific races.)
Because girls have a lower rate of autism, you could expect that girls would reduce the chance of finding significance when the two groups were combined. So , why would you not release the specific data showing the rates of just boys for 2014, 2015, and 2016. It had to have been obtained to get the data on the tables. So why wasn't it included on the 5 pages of tables that you provided for us?
Why would the specific groups that are most likely to be able to prove significance over the three year period only be listed with an overall 2014-2016 combined rate? Why no breakdown year by year? They were that concerned with saving virtual paper?

It comes across as questionable, imo. Would you have designed a study that specifically did not provide a table showing data on the most effected groups by year? Would you yourself, designing a study, have combined girls and boys if you wanted to identify whether an increase in the rate of boys was significant? Maybe this table is somewhere else. In which case, it would be a pretty important one, don't you agree?

By the way, there does appear to be a linear trend. It just hasn't reached the level of 95% significance yet.

In terms of your comment that nothing has changed, so why would the rates have changed, you might want to consider the differences in the immunization schedule over the years.
What recommendations have changed for pregnant women during the time period that would have effected these kids?
Well, in 2006, the CDC strengthened the recommendation for a flu shot in pregnancy. (Kids who would have been approx 8 and under in 2014, 10 and under in 2016 ) By the way, that is a class C risk category shot for pregnant women.
Then in 2011, the TDAP shot in pregnancy after 20 weeks was also recommended, (kids who would be approx 3 years old in 2014, approx 5, 4, and 3 in 2016) Also a class C pregnancy risk category shot.
2008, I think is when it a double dose of influenza was required in a child's first year. ( approx 6 year olds in 2014,)
Many states have also made it much harder even for families who have previously suffered from vaccine injury to refuse shots for their kids.

Also, more and more of the people giving birth will be the first generation of highly vaccinated ( and the first generation who have had to process high injected amounts of aluminium) in the U.S. Will their children already be more susceptible to autoimmune and health problems as a result? We already know environmental toxins given to the mother and father can effect ongoing generations in other animals.

Even if the rates never increased from the current rates or even the 2011 rates, that is still a huge portion of the population, isn't it? Of course given the happy go lucky way they continue to add immune system insults to new babies, regardless of family history, one has to wonder what rates we may be looking at in the future.

I guess this is the leadership persona of the CDC when they have an epidemic that is not easily assigned to vaccination for treatment (or perhaps rather highly suspected if not demonstrated by research at this point to involve a vaccination role).

Imagine a three year survey finding an increase in Guillain-Barre outcomes post influenza-like infections from 1 case for every 45 infections to 1 in 36 and instead of plastering these results all over the corporate media, screaming for increased uptake in the flu shot, pneumonia shots, and calling for funding for several other vaccines in development, they settle on the statistical significance, or lack thereof, in their survey.

Pediatricians waiting for marching orders from on high need to catch on to the fact that they're not going to get appropriate direction from the federal agencies, particularly in this matter, and parents assuming that all this infrastructure protects your babies, need to catch on, too, and probably run from most pediatricians generally, unless willing to risk being enlightened the way most of us here have been enlightened.

RIF

Yes, sampling methods may change but the trend continues to steeple - for instance in Scotland the figure went up from I in 86 in 2011 to 1 in 51 in 2016. These are absolute figures, in the sense that they are number of pupils in the system divided by the number of pupils with an ASD diagnosis, not projections. However, the position is worse than it looks because many children at the younger end will still lack a diagnosis. Indeed, we know the diagnostic and support services are breaking down from the influx. The stories, where they are reported focus perversely on the lack of services and not on the unprecedented scale of the influx. In London, as I was pointing out in May, in 5 boroughs they had services to diagnose 750 children a year but suddenly they had 1500 cases: that could be 10,000 cases a year for the whole of London with a birthrate of 90k. While these extraordinary events happen it is scarcely a defence of the bureaucracy that we don't know all the answers because they employ weak or inconsistent methodology.

Greg

Perhaps, it is the wrong way to look at it. If the CDC wants to find out about something it finds out. If the study method was too blunt to find out whether autism had really risen by half a percent among schoolchildren in two years they are not really trying. If they really wanted to know about it they would find out in 5 minutes about every single case. Ed Yazbak put it memorably in his article eleven years ago "The CDC, Spinach and Autism"

https://www.vaccinationnews.org/f-edward-yazbak-MD-FAAP-articles

The just continue to ruthlessly waste time as the position gets daily worse.

Thanks so much Hera!  Your explanation seems to be in line in with what I was thinking.  A few threads down I also wrote to John discussing the effect that sample size also plays into this.  I consider that the CDC may be guilty of two disingenuous strikes, first, what you're pointing out that they're claiming no significant increase in autism rates over the three years (2014 -2016), and while ignoring the 2013 figures that does indeed suggest a significant increase trend. Second, they fail to address that even for the period from 2014-2016 the finding of no significant increase may have merely been due from the small size.  On this point, let me know if my thinking below is correct.

John, I am not an expert on statistics and I hope someone will correct me if I am thinking this wrong, but I think the CDC is using statistical terminology to confuse and suggest that there was no overall increase in autism rates over those three years, and when that wasn't the case -- there was. Consider that the term 'significant' has a precise statistical definition, which is a 'real' effect that is not due from chance.  Significant results are contingent on sample size.  The larger the sample size, the more likely you are to get a significant result. 

How does this pertain to study?  It means there was an increase, just that the sample size was not large enough to determine whether it was a significant one, not occurring due to  chance.  Yet, this is a far cry from saying there wasn't.  Consider the example where you walk outside and calculate an autism rate for the first 100 kids that you run into at 3 in 100.  Your sample size may be too small to say that that result is significant, but it doesn't mean there is no result.  You could repeat the exercise with a million kids -- a sample size large enough to obtain a significant finding -- and still end up with a rate of 3 autistic kids per 100.

For kids with 'any disability' they were likely dealing with a large enough sample size, and, hence, a significant result was yielded.  Yet, strictly speaking, again, we're not talking about the size of the increase or whether it is greater than the one for increasing autism rates -- but whether it was significant.

Reading is Fundamental

You are confusing yourself. I see no compelling evidence that trend hasn’t continued upwards as it has done every year for the last 30 years. It may be that the data isn’t definitive but it is what it points to - you can make out that it is uncertain but you can’t make out it is level.

Three data points?

Yes, the 2014, 2015, and 2016 NHIS results, unless you want to reanalyze the raw data, in which case you need to understand the original analysis. Note also that the 2014 and 2016 CIs do overlap; it's the sample estimates that don't. The question, if this is what one wants to point at, is formal quantification of significance.

I think a p-value can be extracted from the CIs and SEs by means of an unpaired t-test, but I can't do it on the spur of the moment. You could just E-mail them, as could have Dr. Bob.

Wrong. Of course there is overlap between the three years (there would have to be an almost impossible jump for there not to be) but the 2016 figure is outside the range of 2014, and vica versa.

No, there would not have to be an "almost impossible jump": Note "Other developmental delay" and "Developmental disability." In any event, the point is whether there's a significant linear trend (and there are only three data points). Compare here.

Reading is Fundamental

Wrong. Of course there is overlap between the three years (there would have to be an almost impossible jump for there not to be) but the 2016 figure is outside the range of 2014, and vica versa.

"What I find unusual about the CDC report is that it states 'there was not a statistically significant change in the prevalence of children ever diagnosed with autism spectrum disorder from 2014 to 2016.'"

This again? There wasn't. The data table is sitting right there:

https://www.cdc.gov/nchs/data/databriefs/db291_table.pdf#page=1

The confidence intervals overlap. If Dr. Bob can't figure this out, maybe he should do some remedial work on basic statistics.

This new study on autism prevalence from the CDC is important. Thanks to Dr. Sears for writing about it and to AoA for republishing it.

With 1 in 36 children with an ASD diagnosis, it's obvious that this condition is affecting virtually every classroom and community in the country. And yet, with the issue becoming ever more relevant, the publicity about the escalating rate has all but disappeared. 1 in 500 was news; 1 in 150 was news; 1 in 68 was news; but 1 in 36 is business as usual.

The government and media's apparent lack of concern about rising autism rates has always been shocking to all who read Age of Autism. But the latest silence almost defies comprehension. As Anne Dachel and other instructive writers here so frequently ask, when will the public conversation start? 1 in 20? 1 in 15? 1 in 10? 1 in 5? 1 in 2? Never? It's hard to comprehend.

'What I find unusual about the CDC report is that it states “there was not a statistically significant change in the prevalence of children ever diagnosed with autism spectrum disorder from 2014 to 2016.”

If memory serves me right .. Dr William Thompson .. who has been denied ANY hearing by Congress .. explained how easy it WAS .. (the cynic in me believes it still IS) .. for the CDC to claim there was "no statistically significant change in the prevalence of BLACK children"

Just curious .. since it is claimed .. "for boys, the rate is 1 in 28. For girls, it’s 1 in 80 .. and .. developmental disability increased from 1 in 17 kids to 1 in 14"

If Dr Thompson's information is correct and black children are at "significantly higher risk" of autism .. are any efforts by the CDC to examine if a particular "race" of boys and girls that either confirms or denies Dr. Thompson's allegations?

In other words .. is the rate of 1 in 28 boys the same for black boys" as a group?