Much of modern statistics is anchored in the use of statistics and hypothesis tests that only have desirable and well-known properties when computed from populations that are normally distributed. While it is claimed that many such statistics and hypothesis tests are generally robust with respect to non-normality, other approaches that require an empirical investigation of the underlying population distribution or of the distribution of the statistic are possible and in some instances preferable. In instances when the distribution of a statistic, conceivably a very complicated statistic, is unknown, no recourse to a normal theory approach is available and alternative approaches are required.
Resampling statistics refers to the use of the observed data or of a data generating mechanism (such as a die) to produce new hypothetical samples (resamples) that mimic the underlying population, the results of which can then be analyzed. With numerous cross-disciplinary applications especially in the sub-disciplines of the life science, resampling methods are widely used since they are options when parametric approaches are difficult to employ or otherwise do not apply.
Resampled data is derived using a manual mechanism to simulate many pseudo-trials. These approaches were difficult to utilize prior to 1980s since these methods require many repetitions. With the incorporation of computers, the trials can be simulated in a few minutes and is why these methods have become widely used. The methods that will be discussed are used to make many statistical inferences about the underlying population. The most practical use of resampling methods is to derive confidence intervals and test hypotheses. This is accomplished by drawing simulated samples from the data themselves (resamples) or from a reference distribution based on the data; afterwards, you are able to observe how the statistic of interest in these resamples behaves. Resampling approaches can be used to substitute for traditional statistical (formulaic) approaches or when a traditional approach is difficult to apply. These methods are widely used because their ease of use. They generally require minimal mathematical formulas, needing a small amount of mathematical (algebraic) knowledge. These methods are easy to understand and stray away from choosing an incorrect formula in your diagnostics.
Ecology and Evolution Publicized Applications
This is a list of applications of resampling methods concatenated by Phillip Crowley, 1992.
· Analysis of Null models, competition, and community structure
· Detecting Density Dependence
· Characterizing Spatial Patterns and Processes
· Estimating Population Size and Vital Rates
· Environmental Modeling
· Evolutionary Processes and Rates
· Phylogeny Analysis
In order of use, Crowley found this relationship when searching for publications:
Monte Carlo > Bootstrap > Permutation & Jackknife
Overall, resampling methods were increasing in significant use over the prior decade.
Types of Resampling Methods
I. Monte Carlo Simulation – This is a method that derives data from a mechanism (such as a proportion) that models the process you wish to understand (the population). This produces new samples of simulated data, which can be examined as possible results. After doing many repetitions, Monte Carlo tests produce exact p-values that can be interpreted as an error rate; letting the number of repeats sharpens the critical region.
II. Randomization (Permutation) Test – this is a type of statistical significance test, in which a reference distribution is obtained by calculating all possible values of the test statistic under rearrangements of the labels on the observed data points. Like other the Bootstrap and the Monte Carlo approach, permutation methods for significance testing also produce exact p-values. These tests are the oldest, simplest, and most common form of resampling tests and are suitable whenever the null hypothesis makes all permutations of the observed data equally likely. In this method, data is reassigned randomly without replacement. They are usually based off the Student t and Fisher’s F test. Most non-parametric tests are based on permutations of rank orderings of the data. This method has become practical because of computers; without them, it may be impossible to derive all the possible permutations. This method should be employed when you are dealing with an unknown distribution.
III. Bootstrapping – This approach is based on the fact that all we know about the underlying population is what we derived in our samples. Becoming the most widely used resampling method, it estimates the sampling distribution of an estimator by sampling with replacement from the original estimate, most often with the purpose of deriving robust estimates of standard errors and confidence intervals of a population parameter. Like all Monte Carlo based methods, this approach can be used to define confidence Intervals and in hypothesis testing. This method is beneficial to side step problems with non-normality or if the distribution parameters are unknown. This method can be used to calculate an appropriate sample size for experimental design.
IV. Jackknife – This method is used in statistical inference to estimate the bias and standard error in a statistic, when a random sample of observations is used to calculate it. This method provides a systematic method of resampling with a mild amount of calculations. It offers “improved” estimate of the sample parameter to create less sampling bias. The basic idea behind the jackknife estimator lies in systematically re-computing the statistic estimate leaving out one observation at a time from the sample set. From this new “improved" sample statistic can be used to estimate the bias can be variance of the statistic.
The Flow of Information for resampling methods
In these methods, it is necessary to specify the universe to sample from (random numbers, an observed data set, true or false, etc.), specify the sampling procedure (number of samples, sizes of samples, sampling with or without replacement), and specify the statistic you wish to keep track of. The flow of information is as follows:
1. Input data
2. Resample from the inputted data
3. Calculate the statistic desired
4. Record statistic
5. Return to sample for (X) number of resamples; once reached to completed (X) times, continue to step 6
6. Calculate p-value by counting number of resamples that occur in desired extreme domains divided by the total number of resamples
7. Present/Print results
Resampling Computer Programs
To effectively use these methods, you should have a good program and a fast computer to handle the repetitions. Phillip L. Good has suggested the following programs, with the first four being recommended:
I. R – a programming language that is easy to manipulate. This program is free and scripts are precompiled throughout the Internet. However, be aware, you are on your own.
II. C++ – like R, this is a programming language that has great potential for those entering statistics with a great programming background. Also like R, this is do-it-yourself so you are once again on your own.
III. Resampling Stats – easy to use, this programming language seems very similar to BASIC programming language. It has all the resampling method functions already incorporated and is also available as a Microsoft Excel add-in. It is cheap and easy to follow but can eventually become limited for intense practice of these methods.
IV. S Plus – R based, this program has many built-in functions and pull-down menus, which make it easy to use. The program’s designers offer much support; this package comes at an expensive price.
V. SAS – commonly used in statistical analysis, this package is C based. Pricey and time consuming to debug.
Monte Carlo Simulation
Similar to what was outlined above, the general procedure of Monte Carlo Simulations is as follows:
A. Make a simulated sample population utilizing a non-biased randomizing mechanism (cards, dice, or a computer program) which is based on the population whose behavior you wish to investigate.
B. Create a pseudo-sample to simulate a real-life sample of interest.
C. Repeat step B, (X) number of times.
D. Calculate the probability of interest from the tabulation of outcomes of the resampling trials.
Monte Carlo Example (derived from a Simon 1997 example)
On any given day, it is likely to be sunny 70% of the time. On a sunny day, the Redskins win 65% of their games. What is the likelihood of winning a game on a sunny day? This is a very simplistic example easily described by calculating the joint probability of the two, but serves as a good example of the BASIC programming involved in resampling. The outline:
1.Put seven blue balls and three yellow balls into an urn labeled A (the Nice Day Urn). Put 65 green balls and 35 red balls into an urn labeled B (the Win/Lose Conditional to Nice day Urn).
2.Draw one ball from urn A. If it is blue, continue; otherwise record ‘no’ and stop.
3.If you have drawn a blue ball from run A, now draw a ball from urn B, and if it is green, record ‘yes’; otherwise write ‘no’.
4.Repeat steps 2-3 1000 (or more) times.
5.Count the number of trials.
6.Compute the proportion of ‘yeses’ in the 1000 samples.
The Resampling Statistics programming looks like this:
URN 7#1 3#0 weather Create 10 days samples, 7 nice day
URN 35#0 65#1 winlose Create 100 record sample, 55 wins
REPEAT 1000 Repeat following 1000
SAMPLE 1 weather a Sample 1 of the days
IF a = 1 If a good day, continue; otherwise to skip if statement
SAMPLE 1 winlose b Sample 1 winlose records
IF b = 1 If win, continue; otherwise skip
SCORE b z Tally Good Day Wins
END End ‘if’
END End ‘if’
END Go back to repeat; after 1000, End ‘repeat’
COUNT z = 1 k Count Good Day wins
DIVIDE k 1000 kk Then divide it by 1000
PRINT kk Show on screen the result
Using Resampling Stats® plugin for Microsoft Excel, this is what the print out might look like:
.454 = wins on a sunny day. The actual joint probability approach gives a probability of .455.
Resampling Approaches in Estimation and Hypothesis Testing
I. Hypothesis Testing
A. Normal Theory Approach
For illustration consider Student's t - test for differences in means when variances are unknown, but are considered to be equal. The hypothesis of interest is that
H0: m1 = m2.
While several possible alternative hypotheses could be specified, for our purposes
HA: m1 < m2.
Given two samples drawn from populations 1 and 2, assuming that these are normally distributed populations with equal variances, and that the samples were drawn independently and at random from each population, then a statistic whose distribution is known can be elaborated to test H0:
where are the respective sample means, variances and sample sizes.
When the conditions stated above are strictly met and H0 is true, (1) is distributed as Student's t with (n1 + n2 - 2) degrees of freedom. As
The reasons for making the assumptions specified above is to allow the investigator to make some statement about the likelihood of the computed t - value, and to make a decision as to whether to profess belief in either H0 or Ha. The percentiles of the t distribution with the computed degrees of freedom can be interpreted as the conditional probability of observing the computed t value or one larger (or smaller) given that H0 is true:
Therefore, we would probably wish to profess belief in H0 for suitably large values of a and disbelief for suitably small values. Embedded in this probability we must also include the distributional assumptions mentioned above
If a specific alternative hypothesis had been stated, for example
HA: m1 = m2 - 5,
then under the assumption of normality and equal variances, the t - statistic could be recomputed given the new estimate of m2 under the alternative hypothesis The conditional probability of obtaining the observed difference in t values as computed under the null and alternative hypotheses (t0 - ta), given specified a , the observed variances, and the particular alternative hypothesis could be computed:
This probability is also conditioned on the assumption that both populations are normally distributed
Recall that of these two conditional probabilities a is the Type I error rate, the probability of rejecting the tested null hypothesis when true, and that b is the Type II error rate, the probability of failing to reject the tested null hypothesis when false.
I present this review to emphasize that the estimation of each of these probabilities, which are interpreted as error rates in the process of making a decision about nature, in the course of interpreting a specific statistical test, is totally contingent on assuming specific forms for the distribution of the underlying populations. To know these error rates exactly requires that all the conditions of these tests be met.
B. A distribution-free approach
One way to avoid these distributional assumptions has been the approach now called non - parametric, rank - order, rank - like, and distribution - free statistics. A series of tests many of which apply in situations analogous to normal theory statistics have been elaborated (see ref 1, 2, 3 for expanded treatments of these procedures).
The key to the function of these statistics is that they are based on the ranks of the actual observations in a joint ranking and not on the observations themselves. For example the Wilcoxon distribution - free rank sum test can be applied in place of the 2 sample or separate groups Student's t test. The observations from both samples combined are ranked from least to greatest and the sum of the ranks assigned to the observations from either sample is computed. If both samples are comprised of observations that are of similar magnitude, then the ranks assigned to sample 1 should be similar to the ranks assigned to sample 2. For fixed sample sizes a fixed number of ranks are possible, for n1 = 5 and n2 = 7, 12 ranks will be assigned. Under the null hypothesis that the location of each population on the number line is identical, the sum of the ranks assigned to either sample should equal the sum obtained from randomly assigning ranks to the observations in each sample. In this example there are ways of assigning ranks to sample 1 and or 1 way to assign ranks to sample 2 after assigning the ranks to sample 1. The total number of possible arrangements of the ranks is then and it is possible for each arrangement to compute the sum of the ranks. From this we can enumerate the distribution of the sum of the ranks, W. W in this example can range from 15 to 50. If we divide the
then we have the probability of observing a particular value of w equal to
To obtain the conditional probability that w > W given H0, we simply tabulate the cumulative probabilities.
The only additional assumption embedded in this approach is that the observations are independent, but this is also an assumption of the normal theory approach. Notice we make no assumption about the forms of the underlying populations about which we wish to make inferences, and that the exact distribution of the test statistic, W, is known because it is enumerated. These distribution - free statistics are usually criticized for being less "efficient" than the analogous test based on assuming the populations to be normally distributed. It is true that when the underlying populations are normally distributed then the asymptotic relative efficiencies (ratio of sample sizes of one test to another necessary to have equal power relative to a broad class of alternative hypotheses for fixed a) of distribution - free tests are generally lower than their normal theory analogs, but usually not markedly so. In instances where the underlying populations are non-normal then the distribution - free tests can be infinitely more efficient that their normal theory counterparts. In general, this means that distribution-free tests will have higher Type II error rates (b) than normal theory tests when the normal theory assumptions are met. Type I error rates will not be affected. However, if the underlying populations are not normally distributed then normal theory tests can lead to under estimation of both Type I and Type II error rate.
So far we have used two approaches to estimating error rates in hypothesis testing that either require the assumption of a particular form of the distribution of the underlying population, or that require the investigator to be able to enumerate the distribution of the test statistic when the null hypothesis is true and under specific alternative hypotheses. What can be done when we neither wish to assume normality nor can we enumerate the distribution of the test statistic?
Recall the analogy I used when describing how to generate the expected sum of ranks assigned to a particular sample under the null hypothesis of identical population locations on a number line. The analogy was to a process of randomly assigning ranks to observations independent of one's knowledge of which sample an observation is a member. A randomization test makes use of such a procedure, but does so by operating on the observations rather than the joint ranking of the observations. For this reason, the distribution of an analogous statistic (the sum of the observations in one sample) cannot be easily tabulated, although it is theoretically possible to enumerate such a distribution. From one instance to the next the observations may be of substantially different magnitude so a single tabulation of the probabilities of observing a specific sum of observations could not be made, a different tabulation would be required for each application of the test. A further problem arises if the sample sizes are large. In the example mentioned previously there are only possible arrangements of values so the exact distribution of the sum of observations in one sample could conceivably have been enumerated. Had our sample sizes been 10 and 15 then over 3.2 million arrangements would have been possible. If you have had any experience in combinatorial enumeration then you would know that this approach has rapidly become computationally impractical. With high-speed computers it is certainly possible to tally 3.2 million sums, but developing an efficient algorithm to be sure that each and every arrangement has been included, and included only once is prohibitive.
What then? Sample. When the universe of possible arrangements is too large to enumerate why not sample arrangements from this universe independently and at random? The distribution of the test statistic over this series of samples can then be tabulated, its' mean and variance computed, and the error rate associated with an hypothesis test estimated.
Table 1 contains samples of n1 = 10, and n2 = 15, obtained from sampling from 2 populations with m1 = 200, and m2 = 190, , respectively. A normal theory t - test applied to these data yields a t = 3.3216, df = 23, 0.0005 < p < 0.005. The same data examined by the distribution - free Wilcoxon's rank sum test yields W* = 2.7735, 0.0026 < p < 0.0030. Applying this randomization approach with 1000 iterations of sampling without replacement first 10 and then 15 observations and computing the t statistic for each of these samples yields the distribution depicted in Figure 1. The normal theory t distribution is depicted as a smooth curve. According to the randomization procedure the probability of observing a t value greater than or equal to that actually observed (3.3216) is 0.005 < p < 0.006. Remember that this sampling procedure, unlike an enumeration, allows each possible arrangement of values to be sampled more than once. The probability that on any iteration a particular arrangement will be chosen is in this instance or approximately 3.12 x 10-7. After 1000 such randomizations it is quite possible that some arrangements have been sampled more than once, but there is no reason to believe that particular arrangements yielding either low or high t - values should be systematically included or excluded from the 1000 randomizations.
This approach is obviously an empirical approach to learning something about the distribution of a test statistic under specified conditions. This Monte Carlo sample procedure would have to be performed anew for each new set of observations. One aspect that may be an advantage of this approach over normal theory approaches is that any ad hoc test statistic can be elaborated since a direct empirical investigation of its distributional properties accompanies each test. For example we could just use the difference in the sample means as one test statistic. Figure 2 shows the distribution of over the same 1000 randomizations. The actual difference between sample means is 8.9251 and under the randomization approach the probability of observing a difference this large or larger is 0.004 < p < 0.005. The computed probability of observing the t or actually observed compares favorably with the normal theory estimates.
Figure 3 and 4 illustrate the same procedure applied to two populations whose underlying distributions are exponential. Table 2 presents the sample data generated from two populations with and , respectively. A normal theory test on these data yields t = -2.0874, df = 23, 0.0l <p < 0.025, while the randomization approach yielded a probability of obtaining the observed value of t or one greater of 0.017 < p < 0.018, and a probability of obtaining the observed or a greater difference in means of p > 0.05. Here we see the normal theory test breaking down and convergence in the results obtained from the distribution - free and randomization tests.
Is all of this kosher? We can see the parallel development of the distribution-free and the randomization tests, yet is the randomization test actually yielding a meaningful result? The answer is a resounding well-maybe-er-I-don't-know. The randomization procedure essentially asks the question, given observed samples n1 and n2, if we assume that these samples came from the same underlying population whose distribution F is given by the n1 + n2 sampled values, with probability mass 1/(n1 +n2) for each observation, what is the chance of partitioning the observations into groups of the size observed that have means that differ by an amount as large as that observed? Is this the best empirical estimate of the distribution of a test statistic?
D. The Bootstrap
The Bootstrap is another empirical approach to understanding the distributional properties of a test statistic, but is also useful as a means of estimating statistics and their standard errors. The bootstrap is very similar to the randomization procedure outlined above. The observed distribution of sample values is used as an estimate of the underlying probability distribution of the population F. Then, the distribution of a statistic for fixed sample sizes is obtained by repeatedly sampling from the distribution F, with each value receiving probability mass 1/(n1 + n2), but sampling values with replacement, so that instead of individual partitions of the data having the potential to occur more than once, the individual values themselves may appear repeatedly in a single sample. Under this resampling algorithm the number of possible sample arrangements is much greater than for the randomization approach. For example with a total sample size m = 12, with component samples of size 7 and 5, 127 x 125 = 8.9161004 x 1012 arrangements are possible. For m = 25, and n1 = 10, n2 = 15, 2510 x 2515 = 8.8817842 x 1034 arrangements are possible, factors of 1010 and 1028 more arrangements, respectively, compared to the randomization approach. Any test statistic averaged across a series of say 1000 samples under this algorithm will have a larger standard error since sub-samples of F can deviate from F more than under the randomization algorithm. Figures 5 and 6 illustrate the distribution of t values and for 1000 bootstrap samples of the empirical probability distribution presented in Table 1. For the normal populations the bootstrap estimates the probability of the observed t or one greater to be 0.002 < p < 0.003 which, surprisingly is somewhat less than the randomization approach. This comparison is reversed when examining the differences between means. The bootstrap estimates the probability of the observed mean difference or one greater as 0.022 < p < 0.023, which is an order of magnitude greater than that estimated by the randomization approach, or for that matter for the t -statistic from the same group of bootstrap samples.
Figures 7 and 8 provide similar data for the samples derived from exponentially distributed populations presented in Table 2. The bootstrap is more conservative than either the normal theory approach or the randomization approach when examining the t value obtained for the exponential populations. This is the result I would generally expect in a comparison of the bootstrap and randomization.
Which approach is best? While the randomization approach can be seen to be analogous to the enumeration of distributions that characterizes distribution - free statistics, it is unrealistic in that the distribution of a test statistic across a series of randomized samples is restricted to sub-samples that contain exactly the same observations as the true samples, once each. In some instances this may be the appropriate procedure, but in general randomization may give unrealistically small standard errors for test statistics, so that the true Type I error rates will be greater than nominally stated and Type II error rates also will be greater than nominally stated. However, in all the examples presented above the empirical randomization and bootstrap approaches compare favorably with the normal theory approach.
More examples of Randomization and Bootstrap methods (Simon, 1997):
Simon produced a book “Resampling: the New Statistics”, an example based book on Monte Carlo, Permutation (Randomization) tests, and Bootstrap available for free on the Resampling Stats website. I found the following examples demonstrate the effectiveness of these methods.
Bootstrap example in creating a confidence interval:
Of 135 men of age 34-44 with high cholesterol, 10 developed myocardial infarction. How much confidence should we have that if we were to take a much larger sample than was actually obtained, the sample mean (10/135 = .07) would be in some close vicinity of the actual mean? The general set up may be like this:
1. Construct a sample containing 135 representatives balls: 10 red representing infarction and 125 green representing no infarction
2. Mix, choose a ball, record its color, replace it, and repeat 135 times (to simulate a sample of 135 men).
3. Record the number of red balls among the 135 balls drawn.
4. Repeat steps 2-4 perhaps 1000 times, and observe how much the total number of reds varies from sample to sample.
How this looks in the basic programming:
URN 10#1 125#0 men Create a 135 based sample, 10 of which have myocardial infarction
SAMPLE 135 men a Resample 135 times with replacement
COUNT a =1 b Count how many myocardial infarctions occurred
DIVIDE b 135 c Divide by 135 to get a sample mean
SCORE c z Keep track of every resampled sample mean
HISTOGRAM z Plot the sample means in a histogram
PERCENTILE z (2.5 97.5) k Calculate the 2.5th and 97.5th percentiles
PRINT k Print histogram and results
Using Resampling Stats® plug-in for Microsoft Excel, this is what the print out might look like:
Randomization in hypothesis testing:
The following is the price of whisky in 16 monopoly states (where the state owns the liquor store) and 26 private-owned states are as follows:
16 monopoly states: $4.65, $4.55, $4.11, $4.15, $4.20, $4.55, $3.80,
$4.00, $4.19, $4.75, $4.74, $4.50, $4.10, $4.00, $5.05, $4.20.
Mean = $4.35
26 private-ownership states: $4.82, $5.29, $4.89, $4.95, $4.55, $4.90,
$5.25, $5.30, $4.29, $4.85, $4.54, $4.75, $4.85, $4.85, $4.50, $4.75,
$4.79, $4.85, $4.79, $4.95, $4.95, $4.75, $5.20, $5.10, $4.80, $4.29.
Mean = $4.84
There is a difference in prices of whisky between monopoly and private ownership states with a mean difference of .49. How likely is it that one would choose 16 and 26 states randomly that would have a difference of $.49? To answer this question, the basic programming would look like this:
NUMBERS (4.82 5.29 4.89 4.95 4.55 4.90 5.25 5.30 4.29 4.85 4.54 4.75 Create number array for private which is observed private-ownership sample
4.85 4.85 4.50 4.75 4.79 4.85 4.79 4.95 4.95 4.75 5.20 5.10 4.80 4.29)
NUMBERS (4.65 4.55 4.11 4.15 4.20 4.55 3.80 4.00 4.19 4.75 4.74 4.50 Create number array for monopoly which is observed state owned sample
4.10 4.00 5.05 4.20) mono
CONCAT priv govt all Concatenate the two samples
SHUFFLE 26 all priv$ Allocate 26 of concatenated samples at random w/o replacement to the private resample
SHUFFLE 16 all mono$ Allocate 16 of concatenated samples at random w/o replacement to the monopoly resample
MEAN priv$ p Calculate the mean of the private resample
MEAN mono$ g Calculate the mean of the monopoly resample
SUBTRACT p g diff Subtract the private from monopoly
SCORE diff z Score the difference per resample
END Repeat 1000x then stop
COUNT z >= .49 k Count how many times a resample of equal or greater than .49 occurred; store in k
COUNT z <= -.49 j Count how many times a resample of equal or less than -.49 occurred; store in j
ADD k+j m Add k and j together in n
DIVIDE m/1000 n Calculate a p-value by dividing m by 1000
PRINT m Print result
Using Resampling Stats® plug-in for Microsoft Excel, this is what the print out might look like:
Probability of have a difference greater or equal to .49 is zero. A difference of .49 never occurred in 1000 resamples of the data.
Bootstrap approach in Hypothesis Testing, a difference in treatment
If you believe there is a difference between treatments, act as though they come from different populations; the difference is not due to chance. This is completed by resampling with replacement from individual treatments. The example is as follows: mice that were treated with a certain antibiotic lived 94, 38, 23, 197, 99, 16 and 14 days after their surgery with a mean survival period of 86.8 days. The untreated control lived 52, 10, 40, 104, 51, 27, 146, 30, and 46 days with a mean survival 56.2 days. The difference between the two means is 30.6 days H0: There is a difference in mean survival between treated and untreated.
The basic programming to test this hypothesis might look like this:
NUMBERS (94 38 23 197 99 16 141) treatmt Treatment Sample
NUMBERS (52 10 40 104 51 27 146 30 46) control Control Sample
SAMPLE 7 treatmt treatmt$ Sample 7 treatments w/R.
SAMPLE 9 control control$ Sample 9 controls w/R.
MEAN treatmt$ tmean Mean of Treatments
MEAN control$ cmean Mean of Control
SUBTRACT tmean cmean diff Subtract
SCORE diff scrboard Score the Difference
END Repeat 1000x
HISTOGRAM scrboard Draw a histogram
PERCENTILE scrboard (2.5 97.5) interval Get the 95% Confidence Interval
PRINT interval Give the 95% Confidence Interval
Using Resampling Stats® plug-in for Microsoft Excel, this is what the print out might look like:
The difference of 30.6 days does not suggest that there is a significant difference of post-operational mice that were treated versus untreated.
A. The Jackknife
In the course of applying each of the empirical techniques in the construction of hypothesis tests we could also have estimated test statistics and a suite of characteristics of the test statistics and the empirical distributions. In the test or means we obviously could estimate the means, the variances (or standard errors), and the medians (their standard errors), etc. We could also estimate the bias associated with each of these estimators. Define an estimator of the parameter ,
then the bias of the estimator is c. The sample mean,, is an unbiased estimator of m because , even though different samples may give different estimates of m they are all unbiased estimates. In general, however, most estimators are biased, and the bias can be depicted as a Taylor series expansion of the estimator. So the bias of is
If we define
to be a new estimator of , then the bias of is
which is less than the bias of since it eliminates the term of order 1/n. In practice the estimator is computed as
where i = 1, ... n. This is the first order jackknife estimator. It is useful in that it is a less biased estimator although being somewhat more variable than the un-jackknifed estimator, but this increased variability is at maximum
Since the standard error of an estimator decreases as by a factor of , the estimator has dispersion greater by a factor of 1/n than , but usually only n-3/2 greater than . Therefore the reduction in bias achieved by using is not offset by a similar increase in the magnitude of the estimator's variance. If we depict the bootstrap estimator as then the jackknife estimator of the standard error of is
where FB is the empirical bootstrap probability distribution of the random variables and is a linear approximation of the estimator on the empirical bootstrap probability distribution. This implies that the bootstrap estimator has a standard error that is [n /n - 1]1/2 times less than the jackknife estimator . The jackknife estimator has bias
Where is a quadratic approximation of the estimator on the distribution FB and E indicates the expectation with respect to bootstrap sampling. This implies that the bootstrap estimate of bias is n/(n - 1) times less than the jackknife estimate of bias (see ref. 4).
In general then the bootstrap will provide estimators with less bias and variance than the jackknife. Table 3 shows a data set generated by sampling from two normally distributed populations with m1 = 200, , and m2 = 200 and . To test the hypothesis that the variances of these populations are equal, that is
versus the alternative that
we could use the normal theory approach, which is again conditioned, on the assumptions mentioned earlier and elaborate the test statistic based on the sample estimates of
which is F distributed with n1 - 1 numerator degrees of freedom and n2 - 1 denominator degrees of freedom. Alternatively we could use a jackknife or a bootstrap estimate of the same or a similar test statistic. The F statistic computed under normal theory assumptions is F 9,14 = 8.286, p <0.001, while the bootstrap estimate of the probability of obtaining the observed F or one greater is 0.017 < p < 0.018. The jackknife test is performed on the natural logs of the jackknifed variances rather than the variances themselves. A full description of the computations is given in reference (3). The test statistic for the jackknife test on variances is
where and are the averages of the natural logs of the variances across the n1 and n2 jackknifed estimates and V1 and V2 are the variances of the jackknifed estimates of the variances. For large samples (n1+n2 > 10), Q is N(0,1), but for small equal size samples it follows student's t distribution with n1+n2 - 2 degrees of a freedom. For this example Q = -1.8883, 0.0294 < p < 0.03. Figure 9 presents the distribution of the bootstrap estimates of F, and Table 4 presents the jackknifed pseudo-values their standard errors and bias. Table 5 and Figure 10 provide a similar test for two exponential populations. Under the assumption of normality F9,14 = 2.783, 0.025 < p < 0.05. The jackknife test, however, yields Q = 3.095, 0.0009 < p < 0.001, and the bootstrap yields p > 0.05. In these instances the jackknife is the most powerful test.
III. Recommendations for Use
In 1992, Philip Crowley gave recommendations for these methods:
1. Use a Large Number of Repetitions – use large amount, which will smoothen out the distribution of the data. Use 1000 to start to get a rough idea and use 20,000 as final.
2. In the absence of random sampling, two or more samples with equivalent distributions should be tested by randomization.
3. Don’t use the jackknife approach in confidence intervals and hypothesis testing – randomization and bootstrap approaches are superior.
4. With small sample sizes, be a skeptic of parametric analysis.
Thus far we have presented in a non-rigorous fashion a number of computationally expensive, empirical approaches to estimation and hypothesis testing. The theory underlying some of these approaches is well developed and I refer you to the reference list for that material. However, much of what I have presented has no rigorous theoretical underpinnings, but can be shown to be quite useful particularly in situations where the assumption of normality is suspect. The prognosis among statisticians is that theory will catch up to our computational prowess, so that many of these procedures will be justified and should be adopted. In the interim, however, should you choose to employ one of the more radical of these procedures be prepared for considerable disagreement over its validity and usefulness. The prospects for further development of these kinds of procedures, and work to establish their limitations, advantages, and care and maintenance is considerable. At present, however, the burden of investigating the properties of one of these procedures, in its application to a particular situation and test statistic rests with the investigator.
Bradley, J.U. 1968. Distribution-free statistical tests. Prentice-Hall, Inc: Englewood Cliff, N.J.
Conover, W.J. 1980. Practical Nonparametric statistics. John Wiley and Sons: New York.
Crowley, P. 1992. Resampling Methods for Computation-Intensive Data Analysis in Ecology and Evolution. Annu. Rev. Ecol. Syst. 23:405-447.
Efron, B. and G. Gong.
Good, P.L. 2005. Resampling Methods: 3rd Edition. Birkhauser.
Hollander, M. and D.A. Wolfe. 1973. Nonparametric Statistical Methods. John Wiley and Sons: New York.
Simon, J.L. 1997. Resampling: the New Statistics. Resampling Stats. FREE ONLINE!
Other Readable Literature
Miller, R.G. 1974. The jackknife - a review. Biometrika 61: 1-15.
Peters, S.C. and D.A. Freedman. 1984. Some notes on the Bootstrap in regression problems. Journal of Business and Economic Statistics 2: 406-409.
Efron, B. 1979. Bootstrap Methods: another look at the jackknife. Annals of Statistics 7: 1-26.
Other Not So Readable Literature
Arvesen, J.N. 1969. Jackknifing U-Statistics. Annals of Mathematical Statistics 40: 2076-2100.
Miller, R.G. 1964. A trustworthy jackknife. Annals of Mathematical Statistics 35: 1594-1605.
Miller, R.G. 1968. Jackknifing variances. Annals of Mathematical Statistics 39: 567-582.
Quenouille, M.H. 1956. Notes on bias in estimation. Biometrika 43: 353-360.
Zahl, S. 1977. Jackknifing an index of diversity. Ecology 58: 907-913.
Heltshe, J.F. and N.E. Forrester. 1985. Statistical evaluation of the jackknife estimate of diversity when using quadrat samples. Ecology 66: 107-111.
Routledge, R.D. 1980. Bias in estimating the diversity of large uncensused communities. Ecology 61: 276-281.
Originally by Dr. Edward Connor, modified by Eugenel Espiritu, June 1, 2008.