Monday, March 9, 2015

The frequentist case against the significance test, part 2


The significance test is perhaps the most used statistical procedure in the world, though has never been without its detractors. This is the second of two posts exploring Neyman's frequentist arguments against the significance test; if you have not read Part 1, you should do so before continuing (“The frequentist case against the significance test, part 1”).

Neyman offered two major arguments against the significance test:
  1. The significance test fails as an epistemic procedure. There is no relationship between the \(p\) value and rational belief. More broadly, the goal of statistical inference is tests with good error properties, not beliefs.
  2. The significance test fails as a test. The lack of an alternative means that a significance test can yield arbitrary results.
The first, philosophical, argument I outlined in Part 1. Part 1 was based largely on Neyman's 1957 paper “'Inductive Behavior' as a Basic Concept of Philosophy of Science”. Part 2 will be based on Chapter 1, part 3 of Neyman's 1952 book, “Lectures and conferences on mathematical statistics and probability”.

First, it must be said that Neyman did not think that significance tests were useless or misleading, all the time. He said “The [significance test procedure] has been applied since the invention of the first systematically applied test, the Pearson chi-square of 1900, and has worked, on the whole, satisfactorily. However, now that we have become sophisticated we desire to have a theory of tests.” Obviously, he is not making a blanket statement that significance tests are, generally, good science; he was making an empirical statement about the applications of significance tests in the first half of the twentieth. It is debatable whether he would say the same about the significance test since then.

Of course, we should not evaluate a procedure by its purported results; we can be misled by results, and even worse, this involves an inherent circularity (how do we determine whether the procedure actually performed satisfactorily? Another test?). However, this was merely an informal judgment of Neyman's; we should not over-interpret it either way. After all: he will show that the foundation of the significance test is flawed, and he clearly thought this was important.

An example: Cushney and Peebles' soporific drugs

Suppose that we are interested in the effect of sleep-inducing drugs. Cushney and Peebles (1905) reported the effects of two sleep-inducing drugs on 10 patients in a paired design. Conveniently, R has the data for these 10 patients built-in, as the sleep data set; the data comprise 10 participants' improvements over baseline hours of sleep, for each drug. If we wished to compare the two drugs, we might compute a difference score for each participant and subject these difference scores to a one-sample \(t\) test.

The null hypothesis, in this case, is that the population mean of the difference scores, \(\mu=0\). Making the typical assumptions of normality and independence, we know that, under the null hypothesis,
\[ t = \frac{\bar{x}}{\sqrt{s^2/N}} \sim t_{N-1} \] where \(\bar{x}\) and \(s^2\) are the difference-score sample mean and variance.

The figure below shows the distribution of the \(t\) statistic assuming that the null hypothesis is true, with the corresponding \(p\) values on the top axis. Increasingly red areas show increasing evidence against the null hypothesis, according to the Fisherian view.




If we decided to use the \(t\) statistic to make a decision in a significance test, we would decide on a criterion: say, \(|t|>2.26\), which would lead to \(\alpha=0.05\). Repeating the logic of the significance test, as Neyman put it:
When an “improbable sample” was obtained, the usual way of reasoning was this: “Were the hypothesis \(H\) true, then the probability of getting a value of [test statistic] \(T\) as or more improbable than that actually observed would be (e.g.) \(p = 0.00001\). It follows that if the hypothesis \(H\) be true, what we actually observed would be a miracle. We don't believe in miracles nowadays and therefore we do not believe in \(H\) being true.” (Neyman ,1952)
In the case of our sample, we can perform the \(t\) test in R:
## 
##  Paired t-test
## 
## data:  sleep$extra[1:10] and sleep$extra[11:20]
## t = -4.0621, df = 9, p-value = 0.002833
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -2.4598858 -0.7001142
## sample estimates:
## mean of the differences 
##                   -1.58
In a typical significance test scenario, this would lead to a rejection of the null hypothesis, because \(|t|>2.26\).

Neyman's second argument: Significance testing can be arbitrary

Remember that at this point, we have not considered anything about what we would expect if the null hypothesis were not true. In fact, Fisherian significance testing does not need to consider any alternatives to the null. The pseudo-falsificationist logic of the significance test means that we only need consider the implications for the data under the null hypothesis.

Neyman asks: why use the \(t\) statistic for a significance test? Why use the typical \(\bar{x}\) and \(s^2\)? Neyman then does something very clever: he defines two new statistics, \(\bar{x}_1\) and \(s^2_1\), that have precisely the same distribution as \(\bar{x}\) and \(s^2\) when the null hypothesis is true, and shows that using these two statistics leads to a different test, and different results:
\[ \begin{eqnarray*} \bar{x}_1 &=& \frac{x_1 - x_2}{\sqrt{2N}},\ s^2_1 &=& \frac{\sum_{i=1}^N x_i^2 - N\bar{x}^2_1}{N-1}, \end{eqnarray*} \]
where \(x_i\) is the difference score of the $i$th participant (assuming the samples are in arbitrary order).

Neyman proves that these statistics have the same joint distribution as \(\bar{x}\) and \(s^2\), but we can verify Neyman's proof using R (code available here). The top row of the plot below shows the histogram of 100,000 samples of \(\bar{x}\), \(s^2\), and the \(t\) statistic for \(N=10\) and \(\sigma^2=1\), assuming the null hypothesis is true; the bottom row shows the same 100,000, but computing \(\bar{x}_1\), \(s^2_1\), and \(t_1\), the \(t\) statistic computed from \(\bar{x}_1\) and \(s^2_1\). The red line shows the theoretical distributions. The distributions match precisely.




We now have two sets of statistics that have the same distributions, and will thus produce a significance test with precisely the same properties when the null hypothesis is true. Which should we choose? Fisher might object that \(\bar{x}_1\) and \(s^2_1\) are not sufficient, but this only pushes the problem onto sufficiency: why sufficiency?

The figure below shows that this matters for the example at hand. The figure shows 100,000 simulations of \(t\) and \(t_1\) plotted against one another; when \(t\) is large, \(t_1\) tends to be small; when \(t\) is small, \(t_1\) tends to be large. The red dashed lines show the \(\alpha=0.05\) critical values for each test, and the blue curves show the limits of bounds within which \((t,t_1)\) has to be contained.


The red point shows \(t\) and \(t_1\) for the Cushny and Peebles' data set; \(t\) would lead to a rejection of the null, while \(t_1\) would not.

Examining the definitions of \(\bar{x}_1\) and \(s^2_1\), it isn't difficult to see what is happening; when the null is true, these statistics will have identical distributions to \(\bar{x}\) and \(s^2\). However, when the null is false, they will not. The distribution of \(\bar{x}_1\) will continue to have a mean of 0 (instead of \(\mu\)), while the distribution of \(s^2_1\) will become more spread than \(s^2\). The effect of this is that the power of the test based on \(t_1\) will decrease as the true effect size increases!

A consideration of both Type I and Type II errors makes it obvious which test to choose; we should choose the test that yields the higher power (this is, incidentally, closely related to the Bayesian solution to the problem through the Neyman-Pearson lemma). The use of \(t_1\) would lead to a bad test, when both Type I error rates and Type II error rates are taken into account. A significance test, which does not consider Type II error rates, has no account of why \(t\) is better than \(t_1\).

More problems

The previous development is bad for a significance test; it shows that there can be two statistics that lead to different answers, yet have the same properties from the perspective of significance testing. Following this, Neyman proves something even better: we can always find a statistic that will have the same long-run distribution under the null as \(t\), yet will yield an arbitrarily high test statistic for our sample. This means that we cannot simply base our choice of test statistic on what would yield a more or less conservative test statistic for our sample.

Neyman defines some constants \(\alpha_i\) using the obtained samples \(x_i\): \[ \alpha_i = \frac{x_i}{\sqrt{\sum_{i = 1}^N x_i^2}} \] then for future samples \(y_i\), \(i=1,\ldots,N\) defines \[ \begin{eqnarray*} \bar{y}_2 &=& \frac{\sum_{i=1}^N \alpha_iy_i}{\sqrt{N}},\ s^2_2 &=& \frac{\sum_{i=1}^N y_i^2 - N\bar{y}_2^2}{N-1}, \end{eqnarray*} \] and of course we can compute a \(t\) statistic \(t_2\) based on these values. If we use our observed \(x_i\) values for \(y_i\), this will yield a \(t_2=\infty\), because \(s^2_2 = 0\), exactly! However, if we check the long-run distribution of these statistics under the null hypothesis, we again find that they are exactly the same as \(\bar{x}\), \(s^2\), and \(t\):



If we considered the power of the test based on \(t_2\), we would find that it is worse than the power based on \(t\). The significance test offers no reason why \(t\) is better than \(t_2\), but a consideration of the frequentist properties of the test does. Neyman has thus shown that we must consider an alternative hypothesis in choosing a test statistic, otherwise we can select a test statistic to give us any result we like.

Conclusion: The importance of power

At the risk of belaboring a point that has been made over and over, power is not a mere theoretical concern for a frequentist. Neyman and Pearson offer an account of why some tests are better than others, and also, in some cases, an account of which test is the optimal; however, just because a test is optimal, does not mean it is good.

We might always manage to avoid Type I errors at the same rate (assuming the null hypothesis is true), but as Neyman points out, this is not enough; one needs to consider power, and how one wants to treat both Type I error and power. A good frequentist test may balance Type I and Type II error rates; a good frequentist test may control the Type I error rate while having a power that is above a certain probability. From a frequentist perspective these are decisions that must be made prior to an experiment; none of them can be addressed within the significance testing framework.

To recap both posts, Neyman makes clear why significance testing, as commonly deployed in the scientific literature, does not offer a good theory of inference: it is fails epistemically by allowing arbitrary “rational” beliefs, and it fails on statistical grounds by allowing arbitrary results.

From a frequentist perspective, what might a significance test be useful for? Neyman allows that before a critical set of experiments is performed, exploratory research must be undertaken. Generating a test or a confidence procedure requires some assumptions. Neyman does not offer an account of the process of choosing these assumptions, and seems content to leave this up to substantive researchers. Once a formal inference is needed, however, it is clear that from a frequentist perspective the significance test is inadequate.

[the source to this blog post, including R code to reproduce the figures, is available here: https://gist.github.com/richarddmorey/5806aad7191377dcbf4f]

32 comments:

  1. To me the main problem with confidence interval or hypothesis testing is the following but I don't see discussions such as yours addressing this practical issue :

    Confidence interval or hypothesis testing results depend on the model and the assumptions made. Often there are substantial uncertainties about the model and the assumptions but these are often ignored in performing a hypothesis test. You might prescribe to use robust inference or to put priors on these elements but I don't think these suggestions are often practical.

    On another matter:
    I have tried to use Bayes Factor in a number of occasions but I have found it very sensitive to the priors and in many interesting cases I run into Bartlett paradox.
    A Reader

    ReplyDelete
  2. Bayesian insights are a type of inductive derivation, that is case based surmising. You separate however much data from each case as could be expected and join that data. buy assignment.

    ReplyDelete
  3. We are an exceptionally rumored cleaning organization conveying probably the best cleaning and expert jack of all trades administrations. We are basically outstanding amongst other neighborhood cleaning organizations in Dubai. In addition to the fact that we cater to private customers, we further spotlight on corporate clients too.
    https://www.cleanandshine.ae/

    ReplyDelete
  4. Want to know what are the parts of an essay? Your final essay draft should contain these elements in the following order: title, outline, introduction, several body paragraphs, and conclusion.

    ReplyDelete
  5. experimental articulation about the utilizations of hugeness tests in the main portion of the twentieth. It is easily proven wrong whether he would say the equivalent regarding the centrality test from that point forward.Research Proposal Help

    ReplyDelete
  6. To fabricate a vocation in this field, one ought to have proficient abilities in these three offices: programming, space information and investigation. In programming you have to have a solid information in Python, R, Scala, Julia and Java. You should have the option to comprehend a few expository issues. In SQL, you have to have hands-on understanding. What's more, a short information in Machine Learning.
    office cleaning

    ReplyDelete
  7. A ton of understudies nowadays don't give English the significance it merits and the essential explanation is that they believe that they are very knowledgeable with the language and there is nothing that they could learn. The issue is that learning English sentence structure and upgrading your jargon are two extremely vital necessities all together for an individual to be completely outfitted with the arms stockpile of English.ISO 14001 Saudi Arabia

    ReplyDelete
  8. At the point when you study the subjects like material science, science you get increasingly specialized and inside and out comprehension of the world you live in. It permits you to create logical information and hypothesis about everything on the planet that you go over. Science venture like assignments drive you to consider potential arrangements that can occur inside the setting of this world and even past.
    continental car hire lahore

    ReplyDelete
  9. Understudies who are shown this subject flee from it in light of the fact that the educators may be inept, yet whenever watched, science is an extremely fun subject and empowers the understudy to think carefully control and mental quality. That is the reason it is supposed to be a mandatory subject of primary school so understudies can familiarize themselves with the subject at an early age. It challenges the current standards and customs of the world and causes the understudies to have a positive point of view of the world with a superior comprehension and information.
    Frozen truck for rent in Dubai

    ReplyDelete
  10. Ecommerce website design, cheap website development & Integration company in the UK.Expert PrestaShop, Open Cart, Shopify, Magento and WordPress developers. visit : https://www.budgetweb.ae

    ReplyDelete
  11. Since we realize how to get the entirety of odd arrangements we have to figure what we're including. On the off chance, website design dubai price that we investigate the gatherings we see that for bunch number z, we have z continuous odd digits starting with the [(z^2 - z + 2)/2]th odd digit. This arrangement has z continuous digits so it would end at the [z^2 - z + 2)/2 + z - 1]th odd digit.

    ReplyDelete
  12. These basic advances are the establishments for variable based math, geometry, trigonometry, analytics, direct polynomial math and measurable examination. The more noteworthy, website design dubai price australia dominance you accomplish at the lower levels, the more instinctive you will be at the more elevated levels. When you perceive the examples, math does itself..

    ReplyDelete
  13. Custom Homework Services. It is important to choose the best one among the many. Before choosing the one go for that website reviews first and then check the website like their services.

    ReplyDelete
  14. Rent a car Islamabad,


    We are providing rent a car in Rawalpindi/Islamabad for all your transport needs.When it comes to Car Rental Islamabad, our prime focus is to provide reliable car rental services. Being the best Car Rental Company in Islamabad we provide insured cars

    https://rent-car-islamabad.pk/

    uzairhassan13@yahoo.com

    ReplyDelete
  15. We are a Quality Renovation Contractor that Renews and Brings Life back to Existing Properties regardless . can someone do my assignment

    ReplyDelete
  16. I truly cheerful discovered this site in the long run. Truly educational and defective, Thanks for the post and exertion! It would be ideal if you continue sharing all the more such blog.
    Probability And Statistics Assignment

    ReplyDelete
  17. Drive Cool Transport is one of the distinguished companies that are engaged in offering high quality freezer trucks for rent in UAE. For clients who are engaged in some trading or distribution business, we have the best Freezer Truck Dubai. Our fleet of freezer vehicles is inclusive of Freezer Truck Dubai, Refrigerator Truck Dubai, Refrigerator Van Rental Dubai, Refrigerator Truck With Tail Lift, Refrigerated Truck With Partition, chiller van Van Rental Dubai, Chiller Truck Rental Dubai, and various other types of refrigerated transport.

    ReplyDelete

  18. Special Thanks for Sharing such a good content to aurthor ...But Nowdays rent a car facility available in Islamabad & Rawalpindi One Of My Best Experience with Falcon Rent a car in Islamabad guys you can also avail this opportunity at lowest Rates.
    Rent a car Islamabad services are just one call away. The latest model cars are available for rent in Islamabad with & without driver.
    https://falconrentcar.com/


    ReplyDelete

  19. I was diagnosed as HEPATITIS B carrier in 2013 with fibrosis of the
    liver already present. I started on antiviral medications which
    reduced the viral load initially. After a couple of years the virus
    became resistant. I started on HEPATITIS B Herbal treatment from
    ULTIMATE LIFE CLINIC (www.ultimatelifeclinic.com) in March, 2020. Their
    treatment totally reversed the virus. I did another blood test after
    the 6 months long treatment and tested negative to the virus. Amazing
    treatment! This treatment is a breakthrough for all HBV carriers.

    ReplyDelete
  20. I like your post. It is good to see you verbalize from the heart and clarity on this important subject can be easily observed...Van for rent in Islamabad

    ReplyDelete
  21. We are located in Dubai and provide chiller and freezer van rental service for day, week and month not only in Dubai but also in UAE. We provide our professional and secure services to our valuable customer. For more details please visit our website https://vanchiller.com/

    ReplyDelete
  22. Very good article thank you for sharing energetic information. I really appreciate you.
    keep it up,write and upload on regularly. I read the blog on a daily basis.
    Get the best home tutor in Karachi At your Doorstep. Best Tutor Academy in Karachi For All Classes and Subject


    Jinnah Tutor Academy In Karachi is a trusted tutor academy by parents and children, well-known for its compatibility in a variety of subjects, extensive experienced Home tutor in Karachi to cover loads of curriculum, and determination to guide students on the career-changing path through home tuition.


    We are a team of high-qualified, award-winning, top-rated & professional teachers that not just help students study in crucial hours butHome Tutors also prepare them for future academic challenges.


    ReplyDelete
  23. This comment has been removed by the author.

    ReplyDelete
  24. Afridi refrigerated transport L.L.C (DubaiTruck.com) is a trucking company specializing in chilled & frozen products storage transport & distribution covering Dubai, Abu Dhabi, Sharjah, Ajman, and all over U.A.E. Not only this but we also deal in general transport such as dry boxed and open trucks, vans, and trailers to cover your logistics needs.
    freezer transport in Dubai

    ReplyDelete
  25. In the first place, it should be said that Neyman didn't feel that importance tests were pointless or deceiving, constantly. He said "The [significance test procedure] has been applied since the development of the main deliberately applied test, the Pearson chi-square of 1900, and has worked, in general, acceptably. In any case, since we have become refined we want to have a hypothesis of tests." Obviously, he isn't offering a sweeping expression that importance tests are, by and large, great science; he was offering an experimental expression about the uses of importance tests in the primary portion of the 20th. It is disputable whether he would agree that a similar about the importance test from that point forward. engagement rings

    ReplyDelete
  26. I read your Blog and im really happy with your efforts and I will be waiting for your next write ups thanks once againThanks for sharing your info.

    bedline

    Polyurea

    ReplyDelete
  27. I think you're right that the frequentist case against the significance test isn't really a case of statistics, but rather a case of human nature.

    The problem with the significance test is that it relies on people being able to report their results correctly, and to be honest about whether or not their results are significant. Humans are fallible, and we all make mistakes—it's just human nature. There's no way around this, so we need some way of determining whether or not our results were actually significant (and if they weren't, then why). We may not be able to do this perfectly, but we can do better than just throwing out random data with no regard for its reliability or significance.

    car rentals lahore

    ReplyDelete