Hypothesis Testing with theBootstrapNoa HaasStatistics M.Sc. Seminar, Spring 2017Bootstrap and Resampling Methods
Bootstrap Hypothesis TestingA bootstrap hypothesis test starts with a teststatistic - π‘(π) (not necessary an estimate of aparameter).We seek an achieved significance levelπ΄ππΏ πππππ»0 π‘ π π‘(π)Where the random variable π has a distributionspecified by the null hypothesis π»0 - denote as πΉ0 .Bootstrap hypothesis testing uses a βplug-inβ styleto estimate πΉ0 .
The Two-Sample ProblemWe observe two independent random samples:πΉ πΊ π π§1 , π§2 , , π§π οΏ½οΏ½ πππ π¦1 , π¦2 , , π¦πAnd we wish to test the null hypothesis of no differencebetween F and G,π»0 : πΉ πΊ
Bootstrap Hypothesis Testing πΉ πΊ Denote the combined sample by π, and its empiricaldistribution by πΉ0 . Under π»0 , πΉ0 provides a non parametric estimate forthe common population that gave rise to both π and π.1. Draw π© samples of size π π with replacementfrom π. Call the first n observations π and theremaining π β π 2. Evaluate π‘( ) on each sample - π‘(π π )3. Approximate π΄ππΏππππ‘ byπ΄ππΏππππ‘ # π‘ π π π‘ π /π΅* In the case that large values of π‘(π π ) are evidence against π»0
Bootstrap Hypothesis Testing πΉ πΊon the Mouse DataA histogram of bootstrapreplications ofπ‘ π π§ π¦for testing π»0 : πΉ πΊ on themouse data. The proportion ofvalues greater than 30.63 is .121.Calculatingπ§ π¦π‘ π π 1 π 1 π(approximate pivotal) for thesame replications producedπ΄ππΏππππ‘ .128
Testing Equality of Means Instead of testing π»0 : πΉ πΊ, we wish to test H0 : ππ§ ππ¦ ,without assuming equal variances. We need estimates of πΉand πΊ that use only the assumption of common mean1. Define points π§π π§π π§ π₯, π 1, , π, and π¦π π¦π π¦ π₯ , π 1, , π. The empirical distributions of π and πshares a common mean.2. Draw π© bootstrap samples with replacement π , π fromπ§1 , π§2 , , π§π and π¦1 , π¦2 , , π¦π respectivly3. Evaluate π‘( ) on each sample π¦ π§π‘ π π ππ§ 1 π ππ¦ 1 π4. Approximate π΄ππΏππππ‘ byπ΄ππΏππππ‘ # π‘ π π π‘ π /π΅
Permutation Test VS BootstrapHypothesis Testing Accuracy: In the two-sample problem, π΄ππΏππππ isthe exact probability of obtaining a test statisticas extreme as the one observed. In contrast, thebootstrap explicitly samples from estimatedprobability mechanism. π΄ππΏππππ‘ has nointerpretation as an exact probability. Flexibility: When special symmetry isnβt required,the bootstrap testing can be applied much moregenerally than the permutation test. (Like in thetwo sample problem β permutation test is limitedto π»0 : πΉ πΊ, or in the one-sample problem)
The One-Sample ProblemWe observe a random sample:πΉ π π§1 , π§2 , , π§πAnd we wish to test whether the mean of thepopulation equals to some predetermine value π0 π»0 : ππ§ π0
Bootstrap Hypothesis Testing ππ§ π0What is the appropriate way to estimate the nulldistribution?The empirical distribution πΉ is not anappropriate estimation, because it does notobey π»0 .As before, we can use the empirical distributionof the points:π§π π§π π§ π0 , π 1, , πWhich has a mean of π0 .
Bootstrap Hypothesis Testing ππ§ π0The test will be based on the approximateπ§ π0distribution of the test statistic π‘ π π/ πWe sample π© times π§1 , , π§π with replacementfrom π§1 , , π§π , and for each sample computeπ§ π0 π‘ π π/ πAnd the estimated ASL is given byπ΄ππΏππππ‘ # π‘ π π π‘ π /π΅* In the case that large values of π‘ π π are evidence against π»0
Testing ππ§ π0 on the Mouse DataTaking π0 129, the observed value of the test statistic is86.9 129π‘ π 1.6766.8/ 7(When estimating π with the unbiased estimator for standarddeviation). For 94 of 1000 bootstrap samples, π‘ π wassmaller than -1.67, and thereforπ΄ππΏππππ‘ .094For reference, the studentβs t-test result for the same nullhypothesis on that data gives us42.1π΄ππΏ ππππ π‘6 0.0766.8/ 7
Testing Multimodality of a PopulationA mode is defined to be a local maximum or βbumpβ of thepopulation densityThe data: π₯1 , , π₯485 Mexican stampsβ thickness from 1872.The number of modes is suggestive of the number of distincttype of paper used in the printing.
Testing Multimodality of a PopulationSince the histogram is not smooth, it is difficult to tell from it whether there aremore than one mode.A Gaussian kernel density with window size β estimate can be used in order toobtain a smoother estimate:1π π‘; β πβππ 1π‘ π₯ππβAs π increases, the number of modes in the density estimate is non-increasing
Testing Multimodality of a PopulationThe null hypothesis:π»0 : ππ’ππππ ππ πππππ 1Versus ππ’ππππ ππ πππππ 1. Since the number of modes decreases asβ increases, there is a smallest value of β such that π π‘; β has one mode.Call it β1 . In our case, β1 .0068.
Testing Multimodality of a PopulationIt seems reasonable to use π(π‘; β1 ) as the estimated nulldistribution for our test of π»0 . It is the density estimate thatuses least amount of smoothing among all estimated withone mode (conservative).A small adjustment to π is needed because the formula artificiallyincreases the variance of the estimate with β12 . Let π ; β1 be the rescaleestimate, that imposes variance equal to the sample variance.A natural choice for a test statistic is β1 - a large value of β1 isevidence against π»0 .Putting all of this together, the achieved significance level isπ΄ππΏππππ‘ πππππ ;β1 β1 β1Where each bootstrap sample π is drawn from π ; β1
Testing Multimodality of a PopulationThe sampling from π ; β1 is given by:1 β12 π 2 2π₯π π₯ 1 π¦π π₯ β1 ππ ; π 1, , πWhere π¦1 , , π¦π are sampled with replacement from π₯1 , , π₯π , andππ are standard normal random variables. (called smoothedbootstrap)In the stamps data, out of 500 bootstrap samples, none hadβ1 .0068, so π΄ππΏππππ‘ 0.The results can be interpreted in sequential manner, moving on tohigher values of the least amount of modes. (Silverman 1981)When testing the same for π»0 : ππ’ππππ ππ πππππ 2, 146samples out of 500 had β2 .0033, which translates to π΄ππΏππππ‘ 0.292.In our case, the inference process will end here.
SummaryA bootstrap hypothesis test is carried out using the followings:a) A test statistic π‘(π)b) An approximate null distribution πΉ0 for the data under π»0Given these, we generate π΅ bootstrap values of π‘(π ) under πΉ0 andestimate the achieved significance level byπ΄ππΏππππ‘ # π‘ π π π‘(π) /π΅The choice of test statistic π‘ π and the estimate of the null distributionwill determine the power of the test. In the stamp example, if the actualpopulation density is bimodal, but the Gaussian kernel density does notapproximate it accurately, then the suggested test will not have highpower.Bootstrap tests are useful when the alternative hypothesis is not wellspecified. In cases where there is parametric alternative hypothesis,likelihood or Bayesian methods might be preferable.
Statistics M.Sc. Seminar, Spring 2017 Bootstrap and Resampling Methods . Bootstrap Hypothesis Testing A bootstrap hypothesis test starts with a test statistic - P( ) (not necessary an estimat
May 02, 2018Β Β· D. Program Evaluation ΝThe organization has provided a description of the framework for how each program will be evaluated. The framework should include all the elements below: ΝThe evaluation methods are cost-effective for the organization ΝQuantitative and qualitative data is being collected (at Basics tier, data collection must have begun)
Silat is a combative art of self-defense and survival rooted from Matay archipelago. It was traced at thΓ© early of Langkasuka Kingdom (2nd century CE) till thΓ© reign of Melaka (Malaysia) Sultanate era (13th century). Silat has now evolved to become part of social culture and tradition with thΓ© appearance of a fine physical and spiritual .
On an exceptional basis, Member States may request UNESCO to provide thé candidates with access to thé platform so they can complète thé form by themselves. Thèse requests must be addressed to esd rize unesco. or by 15 A ril 2021 UNESCO will provide thé nomineewith accessto thé platform via their émail address.
ΜΆThe leading indicator of employee engagement is based on the quality of the relationship between employee and supervisor Empower your managers! ΜΆHelp them understand the impact on the organization ΜΆShare important changes, plan options, tasks, and deadlines ΜΆProvide key messages and talking points ΜΆPrepare them to answer employee questions
Dr. Sunita Bharatwal** Dr. Pawan Garga*** Abstract Customer satisfaction is derived from thè functionalities and values, a product or Service can provide. The current study aims to segregate thè dimensions of ordine Service quality and gather insights on its impact on web shopping. The trends of purchases have
ChΓnh VΔn.- CΓ²n Δα»©c ThαΊΏ tΓ΄n thΓ¬ tuα» giΓ‘c cα»±c kα»³ trong sαΊ‘ch 8: hiα»n hΓ nh bαΊ₯t nhα» 9, ΔαΊ‘t ΔαΊΏn vΓ΄ tΖ°α»ng 10, Δα»©ng vΓ o chα» Δα»©ng của cΓ‘c Δα»©c ThαΊΏ tΓ΄n 11, thα» hiα»n tΓnh bΓ¬nh ΔαΊ³ng của cΓ‘c NgΓ i, ΔαΊΏn chα» khΓ΄ng cΓ²n chΖ°α»ng ngαΊ‘i 12, giΓ‘o phΓ‘p khΓ΄ng thα» khuynh ΔαΊ£o, tΓ’m thα»©c khΓ΄ng bα» cαΊ£n trα», cΓ‘i Δược
know how to create bootstrap weights in Stata and R know how to choose parameters of the bootstrap. Survey bootstrap Stas Kolenikov Bootstrap for i.i.d. data Variance estimation for complex surveys Survey bootstraps Software im-plementation Conclusions References Outline
Anatomy of a journal 1. Introduction This short activity will walk you through the different elements which form a Journal. Learning outcomes By the end of the activity you will be able to: Understand what an academic journal is Identify a journal article inside a journal Understand what a peer reviewed journal is 2. What is a journal? Firstly, let's look at a description of a .