|
|
| Line 1: |
Line 1: |
| '''CLs'''<ref name="Read">{{cite journal|last=Read|first=A. L.|title=Presentation of search results: The CL(s) technique|journal=Journal of Physics G: Nuclear and Particle Physics|year=2002|volume=28|issue=10|pages=2693–2704|doi=10.1088/0954-3899/28/10/313|url=http://iopscience.iop.org/0954-3899/28/10/313/}}</ref> (from Confidence Levels) is a [[statistics|statistical]] method for setting upper limits on model [[parameter]]s, a particular form of [[interval estimation]] used for parameters that can take only non-negative values. It was first introduced by physicists working at the [[LEP]] experiment at [[CERN]] and has since been used by many [[high energy physics]] experiments. It is a [[frequentist]] method in the sense that the properties of the limit are defined by means of [[probability of error|error probabilities]], however it differs from standard confidence intervals in that the stated confidence level of the interval is not equal to its [[coverage probability]]. The reason for this deviation is that standard upper limits based on a [[uniformly most powerful test|most powerful test]] necessarily produce empty intervals with some fixed probability when the parameter value is zero, and this property is considered undesirable by most physicists and statisticians.<ref>
| | It"s unarguable that bathroom vanity units are a vital part to any bathroom. Regrettably, many have shied from getting the mirror of these dreams. Why? Well, because bathroom mirror cabinets have become quite costly. Fortuitously, there is an alternative choice for anyone buying the mirror of there dreams on a budget. Welcome to the planet of used bath-room vanities. In the event you choose to learn further on [http://www.iamsport.org/pg/bookmarks/granitedetroitjfy/read/26930526/why-should-you-redesign-your-home bathroom vanity michigan], we recommend many libraries you might consider pursuing. <br><br>Used bath-room mirror cabinets may cost nearly half of what new ones do. This is a major plus side due to the fact most of us only dont have the money to splurge on a new one. Whats more-is that you can obtain a very high-end counter at half the price. If no problems thus they are a genuine bargain used vanities tend to have little. <br><br>So where exactly can you go to get an used bathroom mirror cabinet? A good spot to start could be the web. Head over to your preferred o-nline classifieds site and perform a search for vanities. A number of the popular labeled sites include Recycler.com and Craigslist. Once you search on the site, youre bound to come across many people trying to sell their vanities. We recommend sifting through the outcome at a slow rate and being meticulous in attempting to place one thats truly jaw-dropping. <br><br>Whenever you find the one that meets your likes and budget, contact the manager and setup a meeting. You have to start to see the vanity in-person before shelling out any cash. After the conferences said and done and youre pleased with the counter, go-ahead and purchase it. <br><br>A fast word of caution, when youre shopping for an used counter online make sure to push for a live meeting with owner. Because you werent careful the final thing you want is to have conned for a number of hundred dollars. Browse here at [http://www.feedbooks.com/user/1301025/profile granite kitchen countertops michigan] to read how to allow for it. To get more information, we know you check out: [http://about.me/remodeldetroittyq granite michigan]. All the best and happy mirror shopping!.<br><br>If you loved this write-up and you would like to get much more data regarding [http://www.migente.com/raggedterminolo15 buy health insurance online] kindly stop by our own web-page. |
| {{cite journal|
| |
| author=Mark Mandelkern| title = Setting Confidence Intervals for Bounded Parameters.|journal= Statistical Science|
| |
| volume = 17| number=2|pages= 149–159 | year= 2002 | jstor = 3182816 }}</ref>
| |
| Upper limits derived with the '''CLs''' method always contain the zero value of the parameter and hence the coverage probability at this point is always 100%. The definition of '''CLs''' does not follow from any precise theoretical framework of [[statistical inference]] and is therefore described sometimes as ''ad hoc''. It has however close resemblance to concepts of ''statistical evidence''
| |
| <ref name="Giere">{{cite journal|volume = 36|number = 1|author = Ronald N. Giere|title = Allan Birnbaum's Conception of Statistical Evidence|journal = Synthese|year = 1977|pages = 5–13|url=http://philpapers.org/rec/GIEABC}}</ref>
| |
| proposed by the statistician [[Allan Birnbaum]].
| |
| | |
| == Definition ==
| |
| | |
| Let ''X'' be a [[random sample]] from a [[probability distribution]] with a real non-negative [[parameter]] <math>\theta \in [0,\infty)</math>. A ''CLs'' upper limit for the parameter ''θ'', with confidence level <math>1-\alpha'</math>, is a statistic (i.e., observable [[random variable]]) <math>\theta_{up}(X)</math> which has the property:
| |
| | |
| {{NumBlk|:|<math> \frac{\mathbb{P}( \theta_{up}(X) < \theta | \theta) }{ \mathbb{P}( \theta_{up}(X) < \theta | 0 ) } \leq \alpha' \text{ for all } \theta.</math>|{{EquationRef|1}}}}
| |
| | |
| The inequality is used in the definition to account for cases where the distribution of ''X'' is discrete and an equality can not be achieved precisely. If the distribution of ''X'' is [[Continuous probability distribution|continuous]] then this should be replaced by an equality. Note that the definition implies that the [[coverage probability]] <math>\mathbb{P}( \theta_{up}(X) \geq \theta | \theta)</math> is always larger than <math>1-\alpha'</math>.
| |
| | |
| An equivalent definition can be made by considering a [[hypothesis test]] of the null hypothesis <math>H_0:\theta=\theta_0</math> against the alternative <math>H_1:\theta=0</math>. Then the numerator in ({{EquationNote|1}}), when evaluated at <math>\theta_0</math>, correspond to the [[type I and type II errors|type-I error probability]] (<math>\alpha</math>) of the test (i.e., <math>\theta_0</math> is rejected when <math>\theta_{up}(X) < \theta_0</math>) and the denominator to the [[statistical power|power]] (<math>1-\beta</math>). The criterion for rejecting <math>H_0</math> thus requires that the ratio <math>\alpha/(1-\beta)</math> will be smaller than <math>\alpha'</math>. This can be interpreted intuitively as saying that <math>\theta_0</math> is excluded because it is <math>\alpha'</math> less likely to observe such an extreme outcome as ''X'' when <math>\theta_0</math> is true than it is when the alternative <math>\theta=0</math> is true.
| |
| | |
| The calculation of the upper limit is usually done by constructing a [[test statistic]] <math>q_\theta(X)</math> and finding the value of <math>\theta</math> for which
| |
| | |
| :<math> \frac{\mathbb{P}(q_\theta(X) \geq q_\theta^* |\theta)}{\mathbb{P}( q_\theta(X) \geq q_\theta^* | 0 )} = \alpha' .</math>
| |
| | |
| where <math>q_\theta^*</math> is the observed outcome of the experiment.
| |
| | |
| == Usage in high energy physics ==
| |
| | |
| Upper limits based on the CLs method were used in numerous publications of experimental results obtained at particle accelerator experiments such as [[LEP]], the [[Tevatron]] and the [[LHC]]. Perhaps most notable are the upper limits placed on the production cross section of [[Higgs boson]]s.
| |
| | |
| == Origin ==
| |
| | |
| The original motivation for '''CLs''' was based on a conditional probability calculation suggested by physicist G. Zech
| |
| <ref name="zech">{{cite journal|title = Upper limits in experiments with background or measurement errors| | |
| journal = Nucl. Instrum. and Methods in Phys. Res. Section A|
| |
| volume = 277|
| |
| number = 2-3|
| |
| pages = 608–610|
| |
| year = 1989|
| |
| doi = 10.1016/0168-9002(89)90795-X|
| |
| url = http://www.sciencedirect.com/science/article/pii/016890028990795X|
| |
| author = G. Zech}}</ref> for an event counting experiment. Suppose an experiment consists of measuring <math>n</math> events coming from signal and background processes, both described by [[Poisson distribution]]s with respective rates <math>s</math> and <math>b</math>, namely <math>n \sim \text{Poiss}(s+b)</math>. <math>b</math> is assumed to be known and <math>s</math> is the parameter to be estimated by the experiment. The standard procedure for setting an upper limit on <math>s</math> given an experimental outcome <math>n^*</math> consists of excluding values of <math>s</math> for which <math>\mathbb{P}(n \leq n^*|s+b) \leq \alpha</math>, which guarantees at least <math>1-\alpha</math> coverage. Consider, for example, a case where <math>b=3</math> and <math>n^*=0</math> events are observed, then one finds that <math>s+b \geq 3</math> is excluded at 95% confidence level. But this implies that <math>s \geq 0</math> is excluded, namely all possible values of <math>s</math>. Such a result is difficult to interpret because the experiment cannot essentially distinguish very small values of <math>s</math> from the background-only hypothesis, and thus declaring that such small values are excluded (in favor of the background-only hypothesis) seems inappropriate. To overcome this difficulty Zech suggested conditioning the probability that <math>n \leq n^*</math> on the observation that <math>n_b \leq n^*</math>, where <math>n_b</math> is the (unmeasurable) number of background events. The reasoning behind this is that when <math>n_b</math> is small the procedure is more likely to produce an error (i.e., an interval that does not cover the true value) than when <math>n_b</math> is large, and the distribution of <math>n_b</math> itself is independent of <math>s</math>. That is, not the over-all error probability should be reported but the conditional probability given the knowledge one has on the number of background events in the sample. This conditional probability is easily seen to be
| |
| | |
| :<math>\mathbb{P}(n \leq n^* | n_b \leq n^* , s+b) = \frac{\mathbb{P}(n \leq n^* |s+b)}{\mathbb{P}(n_b \leq n^* |s+b)}
| |
| = \frac{\mathbb{P}(n \leq n^* |s+b)}{\mathbb{P}(n \leq n^* |b)}.</math>
| |
| | |
| which correspond to the above definition of '''CLs'''.
| |
| | |
| === Generalization of the conditional argument ===
| |
| | |
| Zech's conditional argument can be formally extended to the general case. Suppose that <math>q(X)</math> is a [[test statistic]] from which the confidence interval is derived, and let
| |
| | |
| :<math> p_{\theta} = \mathbb{P}( q(X) > q^* | \theta) </math>
| |
| | |
| where <math>q*</math> is the outcome observed by the experiment. Then <math>p_{\theta}</math> can be regarded as an unmeasurable (since <math>\theta</math> is unknown) random variable, whose distribution is uniform between 0 and 1 independent of <math>\theta</math>. If the test is unbiased then the outcome <math>q*</math> implies
| |
| | |
| :<math> p_{\theta} \leq \mathbb{P}( q(X) > q^* | 0 ) \equiv p_0^*</math>
| |
| | |
| from which, similarly to conditioning on <math>n_b</math> in the previous case, one obtains
| |
| | |
| :<math>\mathbb{P}(q(X) \geq q^* | p_\theta \leq p_0^* , \theta) = \frac{\mathbb{P}(q(X) \geq q^* |\theta)}{\mathbb{P}(p_\theta \leq p_0^* |\theta)}
| |
| = \frac{\mathbb{P}(q(X) \geq q^* |\theta)}{p_0^*} = \frac{\mathbb{P}(q(X) \geq q^* |\theta)}{\mathbb{P}( q(X) > q^* | 0 )}.</math>
| |
| | |
| === Relation to foundational principles ===
| |
| | |
| The arguments given above can be viewed as following the spirit of the [[conditionality principle]] of statistical inference, although they express a more generalized notion of conditionality which do not require the existence of an [[ancillary statistic]] . The [[conditionality principle]] however, already in its original more restricted version, formally implies the [[likelihood principle]], a result famously shown by [[Allan Birnbaum|Birnbaum]]<ref>
| |
| {{cite journal|last=Birnbaum|first=Allan|authorlink=Allan Birnbaum|year=1962|title=On the foundations of statistical inference|journal=[[Journal of the American Statistical Association]]|volume=57|issue=298|pages=269–326|doi= 10.2307/2281640|mr=0138176|jstor=2281640 }} ''(With discussion.)''</ref>
| |
| . '''CLs''' does not obey the [[likelihood principle]], and thus such considerations may only be used to suggest plausibility, but not theoretical completeness from the foundational point of view. (The same however can be said on any frequentist method if the [[conditionality principle]] is regarded as necessary).
| |
| | |
| Interestingly, Birnbaum himself suggested in his 1962 paper that the CLs ratio <math>\alpha/(1-\beta)</math> should be used as a measure of the strength of ''statistical evidence'' provided by significance tests, rather than <math>\alpha</math> alone. This followed from a simple application of the [[likelihood principle]]: if the outcome of an experiment is to be only reported in a form of a "accept"/"reject" decision, then the overall procedure is equivalent to an experiment that has only two possible outcomes, with probabilities <math>\alpha</math>,<math>(1-\beta)</math> and <math>1-\alpha</math>,<math>(\beta)</math> under <math>H_1,(H_2)</math>. The [[likelihood ratio]] associated with the outcome "reject <math>H_1</math>" is therefore <math>\alpha/(1-\beta)</math> and hence should determine the evidential interpretation of this result. (Since, for a test of two simple hypotheses, the likelihood ratio is a compact representation of the [[likelihood function]]). On the other hand, if the likelihood principle is to be followed consistently, then the likelihood ratio of the original outcome should be used and not <math>\alpha/(1-\beta)</math>, making the basis of such an interpretation questionable. Birnbaum later described this as having "at most heuristic, but not substantial, value for evidential interpretation".
| |
| | |
| A more direct approach leading to a similar conclusion can be found in Birnbaum's formulation of the ''Confidence principle'', which, unlike the more common version, refers to error probabilities of both kinds. This is stated as follows :<ref>
| |
| {{cite journal|volume = 36|number = 1|last=Birnbaum|first=Allan|authorlink=Allan Birnbaum|title = The Neyman-Pearson Theory as Decision Theory, and as Inference Theory; with a Criticism of the Lindley-Savage Argument for Bayesian Theory|journal = Synthese|year = 1977|pages = 19–49| url=http://philpapers.org/rec/BIRTNT}}</ref>
| |
| | |
| <blockquote>
| |
| "A concept of statistical evidence is not plausible unless it finds
| |
| 'strong evidence for <math>H_2</math> as against <math>H_1</math>' with small probability (<math>\alpha</math>)
| |
| when <math>H_1</math> is true, and with much larger probability (1 -<math>\beta</math>) when
| |
| <math>H_2</math> is true. " </blockquote>
| |
| | |
| Such definition of confidence can naturally seem to be satisfied by the definition of '''CLs'''. It remains true that
| |
| both this and the more common (as associated with the [[Jerzy Neyman|Neyman]]-[[Egon Pearson|Pearson]] theory) versions of the confidence principle are incompatible with the likelihood principle, and therefore no frequentist method can be regarded as a truly complete solution to the problems raised by considering conditional properties of confidence intervals.
| |
| | |
| == Calculation in the large sample limit ==
| |
| If certain regularity conditions are met, then a general likelihood function will become a [[Gaussian function]] in the large sample limit. In such case the '''CLs''' upper limit at confidence level <math>1-\alpha'</math> (derived from the [[uniformly most powerful test]]) is given by<ref name="asimov">{{cite journal|
| |
| author = G. Cowan, K. Cranmer, E. Gross, and O. Vitells |
| |
| title = Asymptotic formulae for likelihood-based tests of new physics |
| |
| journal = Eur.Phys.J. |
| |
| volume = C71 |
| |
| pages = 1554 |
| |
| doi = 10.1140/epjc/s10052-011-1554-0 |
| |
| year = 2011 |
| |
| eprint = 1007.1727 }}</ref>
| |
| | |
| :<math> \theta_{up} = \hat\theta + \sigma\Phi^{-1}(1 - \alpha'\Phi(\hat\theta / \sigma ) ) ,</math>
| |
| | |
| where <math>\Phi</math> is the [[Normal_distribution#Cumulative_distribution_function|standard normal cumulative distribution]], <math>\hat\theta</math> is the [[maximum likelihood]] estimator of <math>\theta</math> and <math>\sigma</math> is its [[standard deviation]]. <math>\sigma</math> might be estimated from the inverse of the [[Fisher information]] matrix or by using the "Asimov"<ref name="asimov"/> data set. This result happens to be equivalent to a [[Bayesian inference|Bayesian]] [[credible interval]] if a uniform [[Prior probability|prior]] for <math>\theta</math> is used.
| |
| | |
| <!--== Criticism ==-->
| |
| | |
| == References ==
| |
| {{Reflist}}
| |
| | |
| ==Further reading==
| |
| * {{cite journal | author = Leon Jay Gleser | title = [Setting Confidence Intervals for Bounded Parameters]: Comment|
| |
| journal = Statistical Science| volume = 17 | number = 2|pages = 161–163 | year = 2002 | jstor = 3182818 }}
| |
| * {{ cite journal|
| |
| journal = Phys. Rev. D | doi = 10.1103/PhysRevD.69.033002 | issue = 3 |
| |
| author = Fraser, D. A. S. and Reid, N. and Wong, A. C. M. |
| |
| title = Inference for bounded parameters |
| |
| year = 2004 |
| |
| url = http://link.aps.org/doi/10.1103/PhysRevD.69.033002 |
| |
| pages = 033002 |
| |
| volume = 69 }}
| |
| * {{cite arXiv|
| |
| author = Robert D. Cousins|
| |
| title = Negatively Biased Relevant Subsets Induced by the Most-Powerful One-Sided Upper Confidence Limits for a Bounded Physical Parameter|eprint = 1109.2023 | year = 2011 }}
| |
| | |
| == External links ==
| |
| * [http://pdg.lbl.gov/2010/reviews/rpp2010-rev-statistics.pdf The Particle Data Group (PDG) review of statistical methods]
| |
| | |
| <!--- Categories --->
| |
| | |
| [[Category:Statistical inference]]
| |
| [[Category:Statistical terminology]]
| |
| [[Category:Measurement]]
| |
| [[Category:Statistical intervals]]
| |
It"s unarguable that bathroom vanity units are a vital part to any bathroom. Regrettably, many have shied from getting the mirror of these dreams. Why? Well, because bathroom mirror cabinets have become quite costly. Fortuitously, there is an alternative choice for anyone buying the mirror of there dreams on a budget. Welcome to the planet of used bath-room vanities. In the event you choose to learn further on bathroom vanity michigan, we recommend many libraries you might consider pursuing.
Used bath-room mirror cabinets may cost nearly half of what new ones do. This is a major plus side due to the fact most of us only dont have the money to splurge on a new one. Whats more-is that you can obtain a very high-end counter at half the price. If no problems thus they are a genuine bargain used vanities tend to have little.
So where exactly can you go to get an used bathroom mirror cabinet? A good spot to start could be the web. Head over to your preferred o-nline classifieds site and perform a search for vanities. A number of the popular labeled sites include Recycler.com and Craigslist. Once you search on the site, youre bound to come across many people trying to sell their vanities. We recommend sifting through the outcome at a slow rate and being meticulous in attempting to place one thats truly jaw-dropping.
Whenever you find the one that meets your likes and budget, contact the manager and setup a meeting. You have to start to see the vanity in-person before shelling out any cash. After the conferences said and done and youre pleased with the counter, go-ahead and purchase it.
A fast word of caution, when youre shopping for an used counter online make sure to push for a live meeting with owner. Because you werent careful the final thing you want is to have conned for a number of hundred dollars. Browse here at granite kitchen countertops michigan to read how to allow for it. To get more information, we know you check out: granite michigan. All the best and happy mirror shopping!.
If you loved this write-up and you would like to get much more data regarding buy health insurance online kindly stop by our own web-page.