A blog on statistics, methods, and open science. Understanding 20% of statistics will improve 80% of your inferences.

Friday, May 20, 2016

Absence of evidence is not evidence of absence: Testing for equivalence

See the follow up post where I introduce my R package and spreadsheet TOSTER to perform TOST equivalence tests, and link to a practical primer on this topic. 

When you find p > 0.05, you did not observe surprising data, assuming there is no true effect. You can often read in the literature how p > 0.05 is interpreted as ‘no effect’ but due to a lack of power the data might not be surprising if there was an effect. In this blog I’ll explain how to test for equivalence, or the lack of a meaningful effect, using equivalence hypothesis testing. I’ve created easy to use R functions that allow you to perform equivalence hypothesis tests. Warning: If you read beyond this paragraph, you will never again be able to write “as predicted, the interaction revealed there was an effect for participants in the experimental condition (p < 0.05) but there was no effect in the control condition (F < 1).” If you prefer the veil of ignorance, here’s a nice site with cute baby animals to spend the next 9 minutes on instead.

Any science that wants to be taken seriously needs to be able to provide support for the null-hypothesis. I often see people switching over from Frequentist statistics when effects are significant, to the use of Bayes Factors to be able to provide support for the null hypothesis. But it is possible to test if there is a lack of an effect using p-values (why no one ever told me this in the 11 years I worked in science is beyond me). It’s as easy as doing a t-test, or more precisely, as doing two t-tests.

The practice of Equivalence Hypothesis Testing (EHT) is used in medicine, for example to test whether a new cheaper drug isn’t worse (or better) than the existing more expensive option. A very simple EHT approach is the ‘two-one-sided t-tests’ (TOST) procedure (Schuirmann, 1987). Its simplicity makes it wonderfully easy to use.

The basic idea of the test is to flip things around: In Equivalence Hypothesis Testing the null hypothesis is that there is a true effect larger than a Smallest Effect Size of Interest (SESOI; Lakens, 2014). That’s right – the null-hypothesis is now that there IS an effect, and we are going to try to reject it (with a p < 0.05). The alternative hypothesis is that the effect is smaller than a SESOI, anywhere in the equivalence range - any effect you think is too small to matter, or too small to feasibly examine. For example, a Cohen’s d of 0.5 is a medium effect, so you might set d = 0.5 as your SESOI, and the equivalence range goes from d = -0.5 to d = 0.5 In the TOST procedure, you first decide upon your SESOI: anything smaller than your smallest effect size of interest is considered smaller than small, and will allow you to reject the null-hypothesis that there is a true effect. You perform two t-tests, one testing if the effect is smaller than the upper bound of the equivalence range, and one testing whether the effect is larger than the lower bound of the equivalence range. Yes, it’s that simple.

Let’s visualize this. Below on the left axis is a scale for the effect size measure Cohen’s d. On the left is a single 90% confidence interval (the crossed circles indicate the endpoints of the 95% confidence interval) with an effect size of d = 0.13. On the right is the equivalence range. It is centered on 0, and ranges from d = -0.5 to d = 0.5.

We see from the 95% confidence interval around d = 0.13 (again, the endpoints of which are indicated by the crossed circles) that the lower bound overlaps with 0. This means the effect (d = 0.13, from an independent t-test with two conditions of 90 participants each) is not statistically different from 0 at an alpha of 5%, and the p-value of the normal NHST is 0.384 (the title provides the exact numbers for the 95% CI around the effect size). But is this effect statistically smaller than my smallest effect size of interest?

Rejecting the presence of a meaningful effect

There are two ways to test the lack of a meaningful effect that yield the same result. The first is to perform two one sided t-tests, testing the observed effect size against the ‘null’ of your SESOI (0.5 and -0.5). These t-tests show the d = 0.13 is significantly larger than d = -0.5, and significantly smaller than d = 0.5. The highest of these two p-values is p = 0.007. We can conclude that there is support for the lack of a meaningful effect (where meaningful is defined by your choice of the SESOI). The second approach (which is easier to visualize) is to calculate a 90% CI around the effect (indicated by the vertical line in the figure), and check whether this 90% CI falls completely within the equivalence range. You can see a line from the upper and lower limit of the 90% CI around d = 0.13 on the left to the equivalence range on the right, and the 90% CI is completely contained within the equivalence range. This means we can reject the null of an effect that is larger than d = 0.5 or smaller than d = -0.5 and conclude this effect is smaller than what we find meaningful (and you’ll be right 95% of the time, in the long run).

[Technical note: The reason we are using a 90% confidence interval, and not a 95% confidence interval, is because the two one-sided tests are completely dependent. You could actually just perform one test, if the effect size is positive against the upper bound of the equivalence range, and if the effect size is negative against the lower bound of the equivalence range. If this one test is significant, so is the other. Therefore, we can use a 90% confidence interval, even though we perform two one-sided tests. This is also why the crossed circles are used to mark the 95% CI.].

So why were we not using these tests in the psychological literature? It’s the same old, same old. Your statistics teacher didn’t tell you about them. SPSS doesn’t allow you to do an equivalence test. Your editors and reviewers were always OK with your statements such as “as predicted, the interaction revealed there was an effect for participants in the experimental condition (p < 0.05) but there was no effect in the control condition (F < 1).” Well, I just ruined that for you. Absence of evidence is not evidence of absence!

We can’t use p > 0.05 as evidence of a lack of an effect. You can switch to Bayesian statistics if you want to support the null, but the default priors are only useful of in research areas where very large effects are examined (e.g., some areas of cognitive psychology), and are not appropriate for most other areas in psychology, so you will have to be able to quantify your prior belief yourself. You can teach yourself how, but there might be researchers who prefer to provide support for the lack of an effect within a Frequentist framework. Given that most people think about the effect size they expect when designing their study, defining the SESOI at this moment is straightforward. After choosing the SESOI, you can even design your study to have sufficient power to reject the presence of a meaningful effect. Controlling your error rates is thus straightforward in equivalence hypothesis tests, while it is not that easy in Bayesian statistics (although it can be done through simulations).

One thing I noticed while reading this literature is that TOST procedures, and power analyses for TOST, are not created to match the way psychologists design studies and think about meaningful effects. In medicine, equivalence is based on the raw data (a decrease of 10% compared to the default medicine), while we are more used to think in terms of standardized effect sizes (correlations or Cohen’s d). Biostatisticians are fine with estimating the pooled standard deviation for a future study when performing power analysis for TOST, but psychologists use standardized effect sizes to perform power analyses. Finally, the packages that exist in R (e.g., equivalence) or the software that does equivalence hypothesis tests (e.g., Minitab, which has TOST for t-tests, but not correlations) requires that you use the raw data. In my experience (Lakens, 2013) researchers find it easier to use their own preferred software to handle their data, and then calculate additional statistics not provided by the software they use by typing in summary statistics in a spreadsheet (means, standard deviations, and sample sizes per condition). So my functions don’t require access to the raw data (which is good for reviewers as well). Finally, the functions make a nice picture such as the one above so you can see what you are doing.

R Functions

I created R functions for TOST for independent t-tests, paired samples t-tests, and correlations, where you can set the equivalence thresholds using Cohen’s d, Cohen’s dz, and r. I adapted the equation for power analysis to be based on d, and I created the equation for power analyses for a paired-sample t-test from scratch because I couldn’t find it in the literature. If it is not obvious: None of this is peer-reviewed (yet), and you should use it at your own risk. I checked the independent andpaired t-test formulas against theresults from Minitab software and reproduced examples in the literature, and checked the power analyses against simulations, and all yielded the expected results, so that’s comforting. On the other hand, I had never heard of equivalence testing until 9 days ago (thanks 'Bum Deggy'), so that’s less comforting I guess. Send me an email if you want to use these formulas for anything serious like a publication. If you find a mistake or misbehaving functions, let me know.

If you load (select and run) the functions (see GitHub gist below), you can perform a TOST by entering the correct numbers and running the single line of code:

TOSTd(d=0.13,n1=90,n2=90,eqbound_d=0.5)

You don’t know how to calculate Cohen’s d in an independent t-test? No problem. Use the means and standard deviations in each group instead, and type:

TOST(m1=0.26,m2=0.0,sd1=2,sd2=2,n1=90,n2=90,eqbound_d=0.5)

You’ll get the figure above, and it calculates Cohen’s d and the 95% CI around the effect size for free. You are welcome. Note that TOST and TOSTd differ slightly (TOST relies on the t-distribution, TOSTd on the z-distribution). If possible, use TOST – but TOSTd (and especially TOSTdpaired) will be very useful for readers of the scientific literature who want quickly check the claim that there is a lack of effect when means or standard deviations are not available. If you prefer to set the equivalence in raw difference scores (e.g., 10% of the mean in the control condition, as is common in medicine) you can use the TOSTraw function.

Are you wondering if your design was well powered? Or do you want to design a study well-powered to reject a meaningful effect? No problem. For an alpha (Type 1 error rate) of 0.05, 80% power (or a beta or Type 2 error rate of 0.2), and a SESOI of 0.4, just type:

powerTOST(alpha=0.05, beta=0.2, eqbound_d=0.4) #Returns n (for each condition)

You will see you need 107 participants in each condition to have 80% power to reject an effect larger than d = 0.4, and accept the null (or an effect smaller than your smallest effect size of interest). Note that this function is based on the z-distribution, it does not use the iterative approach based on the t-distribution that would make it exact – so it is an approximation but should work well enough in practice.

TOSTr will perform these calculations for correlations, and TOSTdpaired will allow you to use Cohen’s dz to perform these calculations for within designs. powerTOSTpaired can be used when designing within subject design studies well-powered to test if data is in line with the lack of a meaningful effect.

Choosing your SESOI

How should you choose your SESOI? Let me quote myself (Lakens, 2014, p. 707):
In applied research, practical limitations of the SESOI can often be determined on the basis of a cost–benefit analysis. For example, if an intervention costs more money than it saves, the effect size is too small to be of practical significance. In theoretical research, the SESOI might be determined by a theoretical model that is detailed enough to make falsifiable predictions about the hypothesized size of effects. Such theoretical models are rare, and therefore, researchers often state that they are interested in any effect size that is reliably different from zero. Even so, because you can only reliably examine an effect that your study is adequately powered to observe, researchers are always limited by the practical limitation of the number of participants that are willing to participate in their experiment or the number of observations they have the resources to collect.

Let’s say you collect 50 participants in two independent conditions, and plan to do a t-test with an alpha of 0.05. You have 80% power to detect an effect with a Cohen’s d of 0.57. To have 80% power to reject an effect of d = 0.57 or larger in TOST you would need 66 participants in each condition.

Let’s say your SESOI is actually d = 0.35. To have 80% power in TOST you would need 169 participants in each condition (you’d need 130 participants in each condition to have 80% power to reject the null of d = 0 in NHST).

Conclusion

We see you always need a bit more people to reject a meaningful effect, than to reject the null for the same meaningful effect. Remember that since TOST can be performed based on Cohen’s d, you can use it in meta-analyses as well (Rogers, Howard, & Vessey, 1993). This is a great place to use EHT and reject a small effect (e.g., d = 0.2, or even d = 0.1), for which you need quite a lot of observations (i.e., 517, or even 2069).

Equivalence testing has many benefits. It fixes the dichotomous nature of NHST. You can now 1) reject the null, and fail to reject the null of equivalence (there is probably something, of the size you find meaningful), 2) reject the null, and reject the null of equivalence (there is something, but it is not large enough to be meaningful, 3) fail to reject the null, and reject the null of equivalence (the effect is smaller than anything you find meaningful), and 4) fail to reject the null, and fail to reject the null of equivalence (undetermined: you don’t have enough data to say there is an effect, and you don’t have enough data to say there is a lack of a meaningful effect). These four situations are visualized below.

There are several papers throughout the scientific disciplines telling us to use equivalence testing. I’m definitely not the first. But in my experience, the trick to get people to use better statistical approaches is to make it easy to do so. I’ll work on a manuscript that tries to make these tests easy to use (if you read this post this far, and work for a journal that might be interested in this, drop me a line – I’ll throw in an easy to use spreadsheet just for you). Thinking about meaningful effects in terms of standardized effect sizes and being able to perform these test based on summary statistics might just do the trick. Try it. 








Lakens, D. (2013). Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs. Frontiers in Psychology, 4. http://doi.org/10.3389/fpsyg.2013.00863

Lakens, D. (2014). Performing high-powered studies efficiently with sequential analyses: Sequential analyses. European Journal of Social Psychology, 44(7), 701–710. http://doi.org/10.1002/ejsp.2023

Rogers, J. L., Howard, K. I., & Vessey, J. T. (1993). Using significance tests to evaluate equivalence between two experimental groups. Psychological Bulletin, 113(3), 553.

Schuirmann, D. J. (1987). A comparison of the two one-sided tests procedure and the power approach for assessing the equivalence of average bioavailability. Journal of Pharmacokinetics and Biopharmaceutics, 15(6), 657–680.


The R-code that was part of this blog post is temporarily unavailable as I move it to a formal R package.

25 comments:

  1. Thank you for this nice post. An analogous procedure exists in Bayesian estimation: Check whether the posterior credible interval falls within the SESOI. I like the Bayesian version better than the frequentist version because frequentist confidence intervals change when the stopping or testing intentions change, but Bayesian intervals don't depend on those intentions. If interested, see Ch 12 of DBDA2E, or pp. 16-17 of this manuscript.

    ReplyDelete
    Replies
    1. It's not "intentions" that change but rather the relevant error probabilities (as a result of things like optional stopping, cherry picking, biasing selection effects).

      https://errorstatistics.com/2015/05/27/intentions-is-the-new-code-word-for-error-probabilities-allan-birnbaums-birthday/

      It's very strange that users of tests wouldn't know how to interpret insignificant results when it's part of N-P testing. I'm curous as to how this use of equivalence testing compares with (a) power analysis and (b) a severity analysis of a negative result. See for example section 3.1, 4.2 and 4.3 of Mayo and Spanos (for a one sample Normal test). A severity analysis doesn't require that you set a range of interest or equivalence.
      http://www.phil.vt.edu/dmayo/personal_website/2006Mayo_Spanos_severe_testing.pdf

      Delete
    2. As far as I know, Neyman recommends to interpret a p > 0.05 as either accepting the null, or 'remaining in doubt'. It seems to me that equivalence is a nice way to differentiate between the two depending on the smallest effect size of interest (and assuming you did not have 99% power for that effect size). I don't know how it is related to a severity test - sounds like a useful blog post on your end!

      Delete
  2. Very nice post. I remember we had this in a courses I TA like 7 seven years ago (I was undergrad then) but I also remember I didn't really see the point over simply eyeballing the CI. And then I totally forgot about this until you brought it up.

    (The reason why we had it was probably that they had minitab as well as SPSS! But I honestly don't remember )


    What is in your opinion the benefits over simply calculating a 95 CI around the observed d? I can think of two 1) eyeballing is not very precise 2) p can be used as a continuous measures in an easier fashion. Are there any others? Am I missing something completely??

    ReplyDelete
  3. Similar to Rickard, I had equivalence testing in stats intro course ca. 10 years ago :)

    What if your equivalence test fails to reject the equivalence hypothesis? Would you perform a post-hoc test for significant difference? Isn't this HARKing?

    To be sure bayes factors can't avoid the inferential limbo either if the evidence isn't decisive (BF~1). But they at least can separaty "non-sig difference due to small power" (BF~1) from the lack of difference (BF_01<1).

    I recall reading about frequentist three-way hypothesis tests (reject H1, reject H0, needs more power), but I haven't seen them in use.

    ReplyDelete
    Replies
    1. You'd have one of the two situations in the 4 graphs at the end of the post, right? So either a significant meaningful effect, or an undetermined situation. As far as I understand you can perform both tests (NHST and EHT), and you interpret them both. And it seems stat training was a bit more complete where you had it than where I had it 10 years ago :)

      Delete
  4. Can you explain in nice small words why the equivalence range goes from -0.5 to +0.5, rather than from 0 to +0.5 or perhaps minus infinity to +0.5? That seems to imply that I'm equally interested in results in both directions. But if I'm testing medicines, for example, I don't really care (i.e., I don't have to distinguish between) whether my new pill is less good than the old one, or no good at all, or kills people; I just want to know if it's better than the old one.

    Maybe what I'm saying is, this all sounds a bit two-tailed, so how would it fit into a one-tailed world? Or (most likely) have I missed something?

    ReplyDelete
    Replies
    1. Hi Nick - yes, all that is possible (and makes sense). You can test for noninferiority, for example. You can also set the equivalence range any way you like (from -0.1 to 0.5). The symmetric situation is easiest, my code only works with symmetrical intervals (but I can update it). I discussed it in an earlier draft, but the blog was already so long, I removed it. But you know I am a big fan of one-sided tests if you have one-sided hypotheses, and that generalizes to equivalence tests.

      Delete
    2. Thanks. Looks like I actually understood for once. :-) Your previous post(s) about one-tailed tests were a big part of why I asked.

      Delete
  5. Thanks Nick for pointing out the very useful equivalence test. Perhaps you are not aware of the 'equivalence' R package, but if you are, how does your implementation differ?

    ReplyDelete
    Replies
    1. Hi Remko, from my blog post above:

      One thing I noticed while reading this literature is that TOST procedures, and power analyses for TOST, are not created to match the way psychologists design studies and think about meaningful effects. In medicine, equivalence is based on the raw data (a decrease of 10% compared to the default medicine), while we are more used to think in terms of standardized effect sizes (correlations or Cohen’s d). Biostatisticians are fine with estimating the pooled standard deviation for a future study when performing power analysis for TOST, but psychologists use standardized effect sizes to perform power analyses. Finally, the packages that exist in R (e.g., equivalence) or the software that does equivalence hypothesis tests (e.g., Minitab, which has TOST for t-tests, but not correlations) requires that you use the raw data. In my experience (Lakens, 2013) researchers find it easier to use their own preferred software to handle their data, and then calculate additional statistics not provided by the software they use by typing in summary statistics in a spreadsheet (means, standard deviations, and sample sizes per condition). So my functions don’t require access to the raw data (which is good for reviewers as well). Finally, the functions make a nice picture such as the one above so you can see what you are doing.

      Delete
  6. Hi Daniel. For good reason, Popper's principle of falsifiability has been a pillar of science but even Gosset and Fisher recognised imperfections of 0.05 as a cut off. As you state, over- and under-powered tests will mask meaningful effects. In medicine and sport, there are two key questions about interventions: first, does the treatment/training work and second, if yes, how well? With equivalence-type trials where effects of a new therapy are compared with those of usual care, the conventional null hypothesis testing approach can, perhaps, be retained via the minimum clinically (or practically) important difference that is declared at the outset and that must be exceeded before the new treatment can be considered to be an improvement. The next stage is to evaluate if the improvement is cost effective. Apologies if I have missed something but that wasn't clear in your otherwise helpful account. Incidentally the there-was-no-effect-(P > 0.05)-but-oh-yes-there-was-(d = 0.36) is the pantomime that arises from mixing null-hypothesis significance testing and magnitude-based inferences. Especially when alpha (0.05) is stated in the methods section. The authors of such a statement are using oleaginous Uriah-Heap statements to cover their backs but in fact, by so doing, confuse both themselves and readers.

    ReplyDelete
    Replies
    1. Hi, if you set a smallest effect size of interest, poewr up for it, and don't find a significant result, you might but don't automatically, have evidence for an effect SMALLER than your SESOI. You could be in the 'undetermined' condition visualized above.

      Delete
  7. Hi Daniels, interesting stuff! A couple more or less random thoughts on this:

    1) Equivalence testing is usually used only when a researcher actually hypothesises that a particular null is true. But it can be used more widely: there's no reason we couldn't generally approach inference about any parameter as a problem of working out whether we can conclude that a parameter is trivially small, conclude that it is reasonably large, or conclude that there is too much uncertainty to say. We can do that by combining equivalence testing with traditional NHST, or using something like magnitude-based inference as used in sports science - see http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0147311

    2) The frequentist approach works fine, but one key advantage of a Bayesian approach here is that we can take into account the fact that most effects in psychology are small. I.e., we can place a prior that represents a belief that most effects aren't too far from zero. That makes it less likely that we'll conclude there's a substantial effect, but also more likely that we can conclude confidently that an effect is trivial.

    ReplyDelete
  8. Hi, Daniel. Thanks for a very interesting post and for making your R code available.

    In the conclusion, you refer to the "null of equivalence." Strictly speaking, shouldn't this be the null hypothesis of nonequivalence?

    See Rogers, Howard, and Vessey (1993, p. 554): "There is a null hypothesis asserting that the difference between the two groups is at least as large as the one specified by the investigator [i.e., nonequivalence], and there is an alternative hypothesis asserting that the difference between two groups is smaller than the specified one [i.e., equivalence]."

    ReplyDelete
    Replies
    1. Yes, the null of non-equivalence or the null in equivalence tests is correct, the null of equivalence is not correct.

      Delete
  9. Hi Daniel,
    thanks for making it so easy to conduct these tests.

    Are you planning on amending the syntax to provide a power analysis for TOST r (correlations)?
    That would be really useful for me.

    ReplyDelete
    Replies
    1. Hi, I'm thinking about turning this into a paper - will look into a power analysis for r script, indeed, makes sense to provide!

      Delete
  10. If you only do an equivalence test after p > 0.05 isn't alpha now inflated for that test?

    ReplyDelete
  11. I'm late to the game...

    I agree with nearly everything written in the post, except one I believe crucial issue.

    In the section under "Rejecting the presence of a meaningful effect" it reads:
    "This means we can reject the null of an effect that is larger than d = 0.5 or smaller than d = -0.5 and conclude this effect is smaller than what we find meaningful (and you’ll be right 95% of the time, in the long run)."

    The first part of the sentence is of course correct, But the second part makes a probabilistic claim about the truth of the alternative hypothesis, which cannot be made in the frequentist framework (yes, the sentence uses frequentist language, but the inference is about the truth of the hypothesis). If one wants to make such claims, one would need to use Bayes and a prior to go from P(data|H0) to p(H1|data).

    I think a more accurate version of the cited sentence would be

    "This means we can reject the null of an effect that is larger than d = 0.5 or smaller than d = -0.5
    because the probability of the observed data given the hypothesis that |d| > .05 is smaller than 5%."

    Maybe that doesn't sound very satisfying, but if one likes to make statements about the probability of hypotheses there is no way around a Bayesian approach.

    ReplyDelete
  12. Applying for an admission or scholarship to any national or international university requires a person to submit either filled hard copy or an online application form. Attaching certified documents, financial statement, English Proficiency test result and 3 recommendation letters are the pre-mandatory items to be submitted along with an application form. See more programming homework

    ReplyDelete
  13. Hi,
    I probably missed this, but wouldn't you need a Bonferroni type of correction when looking at two tests?

    ReplyDelete
    Replies
    1. Very good question. I would think so as well, but I've not found a reference that does the simulations to show this. I will do them and let you know, when I publish a paper about this.

      Delete
  14. Have you had experience using XLSTAT 'add on' software for Excel to calculate TOST? Their online tutorial makes it look simple. I am unfamiliar with R and to save me learning it, I thought this might be useful for equivalence testing. Any thoughts? Many thanks in advance.

    ReplyDelete
    Replies
    1. Never heard about it, but why don't you just use the soreadsheet that comes with my 2017 article? And R is SUPER easy if you just want to use TOST. Like a simple calculator.

      Delete