Frames2: a Package for Estimation in Dual Frame Surveys

Data from complex survey designs require special consideration with regard to estimation of finite population parameters and corresponding variance estimation procedures, as a consequence of significant departures from the simple random sampling assumption. In the past decade a number of statistical software packages have been developed to facilitate the analysis of complex survey data. All these statistical software package are able to treat samples selected from one sampling frame containing all population units. Dual frame surveys are very useful when it is not possible to guarantee a complete coverage of the target population and may result in considerable cost savings over a single frame design with comparable precision. There are several estimators available in the statistical literature but no existing software covers dual frame estimation procedures. This gap is now filled by package Frames2. In this paper we highlight the main features of the package. The package includes the main estimators in dual frame surveys and also provides interval confidence estimation.


Introduction
Classic sampling theory usually assumes the existence of one sampling frame containing all finite population units.Then, a probability sample is drawn according to a given sampling design and information collected is used for estimation and inference purposes.In traditional 'design-based' inference the population data are regarded as fixed and the randomness comes entirely from the sampling procedure.The most used design-based estimator is the Horvitz-Thompson estimator that is unbiased for the population total if the sampling frame includes all population units, if all sampled units respond and if there is no measurement error.In the presence of auxiliary information, there exist several procedures to obtain more efficient estimators for population means and totals of variable of interest; in particular, customary ratio, regression, raking, post-stratified and calibration estimators.Several software packages have been developed to facilitate the analysis of complex survey data and implement some of these estimators as SAS (SAS Institute Inc., 2013), SPSS (IBM Corporation, 2013), Systat (Systat Software Inc., 2009), Stata (Stata Corporation, 2015), SUDAAN (Research Triangle Institute, 2013) or PCCarp (Fuller et al., 1989).CRAN hosts several R packages that include these design-based methods typically used in survey methodology to treat samples selected from one sampling frame (e.g.survey, Lumley 2014; sampling, Tillé and Matei 2012; laeken, Alfons et al. 2014 or TeachingSampling, Gutierrez Rojas 2014, among others).Templ (2014) provides a detailed list of packages that includes methods to analyse complex surveys.
In practice, the assumption that the sampling frame contains all population units is rarely met.Often, one finds that sampling from a frame which is known to cover approximately all units in the population is quite expensive while other frames (e.g.special lists of units) are available for cheaper sampling methods.However, the latter usually only cover an unknown or only approximately known fraction of the population.A common example of frame undercoverage is provided by telephone surveys.Estimation could be affected by serious bias due to the lack of a telephone in some households and the generalised use of mobile phones, which are sometimes replacing fixed (land) lines entirely.The potential for coverage error as a result of the exponential growth of the cell-phone only population has led to the development of dual-frame surveys.In these designs, a traditional sample from the landline frame is supplemented with an independent sample from a register of cell-phone numbers.
The dual frame sampling approach assumes that two frames are available for sampling and that, overall, they cover the entire target population.The most common situation is the one represented in Figure 1 where the two frames, say frame A and frame B, show a certain degree of overlapping, so it is possible to distinguish three disjoint non-empty domains: domain a, containing units belonging to frame A but not to frame B; domain b, containing units belonging to frame B but not to frame A and domain ab, containing units belonging to both frames.As an example, consider a telephone survey where both landline and cell phone lists are available; let A be the landline frame and B the cell phone frame.Then, it is possible to distinguish three types of individuals: landline only units, cell-only units and units with both landline and cell phone, which will compose domain a, b and ab, respectively.
Nevertheless, one can face some other situations depending on the relative positions of the frames.For example, Figure 2 shows the case in which frame B is totally included in frame A, that is, frame B is a subset of frame A.Here domain b is empty.We also may find scenarios where the two sampling frames exactly match, as depicted in Figure 3, where ab is the only non-empty domain.Finally, the scenario where domain ab is empty has no interest from a dual frame perspective, since it can be The R Journal Vol.XX/YY, AAAA 20ZZ ISSN 2073-4859   considered as a special case of stratified sampling.
Whatever the scenario, an appropriate choice of the frames results in a better coverage of the target population, which, in turn, leads to a better efficiency of estimators calculated from data from dual frame surveys.This point is particularly important when estimating parameters in rare or elusive populations, where undercoverage errors are usually due to the difficulty of finding individuals showing the characteristic under study when sampling from only one general frame.This issue can be dealt with by incorporating a second frame with a high density of members of the rare population so that the two frames are, together, now complete.Dual frame sampling as a method of improvement of efficiency may seem expensive and unviable, but it is not.In fact, Hartley (1962) notes that dual frame surveys can result in important cost savings in comparison with single frame ones with a comparable efficiency.As an additional interesting characteristic, dual frame methodology offers the researcher the possibility to consider different data collection procedures and/or different sampling designs, one for each frame.Dual frame surveys have gained much attention and became largely used by statistical agencies and private organizations to take advantage of these benefits.
Standard software packages for complex surveys cannot be used directly when the sample is obtained from a dual frame survey because the classical design-based estimators are severely biased and there is a underestimation of standard errors.Weighted analyses with standard statistical software, with certain modified weights, can yield correct point estimates of population parameters but still yield incorrect results for estimated standard errors.A number of authors have developed methods for estimating population means and totals from dual (or, more generally, multiple) frame surveys but most of these methods require ad-hoc software for their implementation.To the best of our knowledge, there is no software incorporating these estimation procedures for handling dual frame surveys.
Frames2 (Arcos et al., 2015) tries to fill this gap by providing functions for point and interval estimation from dual frame surveys.The paper is organized as follows.In the next section, we provide an overview of the main point estimators proposed so far in the dual frame context and review also jackknife variance estimation as a tool to compare efficiency for all of them.Subsequently, we present package Frames2, discussing guidelines that have been followed to construct it and presenting its principal functions and functionalities.We also provide examples to illustrate how the package works.

Estimation in dual frame surveys
Consider again the situation depicted in Figure 1.Assume we have a finite set of N population units identified by integers, U = {1, . . ., k, . . ., N}, and let A and B be two sampling frames, both can be incomplete, but it is assumed that together they cover the entire finite population.
Let A be the set of population units in frame A and B the set of population units in frame B. The population of interest, U , may be divided into three mutually exclusive domains, and N ab be the number of population units in U , A, B, a, b, ab, respectively.
Let y be a variable of interest in the population and let y k be its value on unit k, for k = 1, . . ., N. The objective is to estimate the finite population total Y = ∑ k y k that can be written as where Y a = ∑ k∈a y k , Y ab = ∑ k∈ab y k and Y b = ∑ k∈b y k .To this end, independent samples s A and s B are drawn from frame A and frame B of sizes n A and n B , respectively.Unit k in A has first-order inclusion probability π A k = Pr(k ∈ s A ) and unit k in B has first-order inclusion probability π B k = Pr(k ∈ s B ). From data collected in s A , it is possible to compute one unbiased estimator of the total for each domain in frame A, Ŷa and ŶA ab , as described below: The R Journal Vol.XX/YY, AAAA 20ZZ ISSN 2073-4859 where δ k (a) = 1 if k ∈ a and 0 otherwise, δ k (ab) = 1 if k ∈ ab and 0 otherwise and d A k are the weights under the sampling design used in frame A, defined as the inverse of the first order inclusion probabilities, d A k = 1/π A k .Similarly, using information included in s B , one can obtain an unbiased estimator of the total for domain b and another one for domain ab, Ŷb and ŶB ab , which can be expressed as and 0 otherwise, and d B k the weights under the sampling design used in frame B defined as the inverse of the first order inclusion probabilities, d B k = 1/π B k .Different approaches for estimating the population total from dual frame surveys have been proposed in the literature.Hartley (1962) suggests the use of a parameter, θ, to weight ŶA ab and ŶB ab , providing the estimator where θ ∈ [0, 1].Hartley (1974) himself proved that is the optimum value for θ so that variance of the estimator with respect to the design is minimized.
In practice, θ opt cannot be computed, since population variances and covariances involved in its calculation are unknown, so they must be estimated from sampling data.An estimator for the variance of ŶH can be computed, taking into account that samples from frame A and frame B are drawn independently, as follows where hats denote suitable variance and covariance estimators.Fuller and Burmeister (1972) introduce information from the estimation of overlap domain size, obtaining the following estimator are the optimal values for β 1 and β 2 in the sense that they minimize the variance of the estimator.Again, β1 and β2 need to be estimated, since population values are not known in practice.An estimator for the variance of ŶFB is given by with ŶB = Ŷb + ŶB ab .Bankier (1986) and Kalton and Anderson (1986) combine all sampling units coming from the two frames, s A and s B , trying to build a single sample as if it was drawn from only one frame.Sampling weights for the units in the overlap domain need, then, to be modified to avoid bias.These adjusted weights are (5) Hence, the estimator can be expressed in the form The R Journal Vol.XX/YY, AAAA 20ZZ ISSN 2073-4859 with s = s A ∪ s B .Note that to compute this estimator, one needs to know, for units in the sample coming from the overlap domain, the inclusion probability under both sampling designs.Rao and Skinner (1996) propose the following unbiased estimator for the variance of the estimator where zA When frame sizes, N A and N B , are known, estimator (6) can be adjusted to increase efficiency through different procedures as, for example, raking ratio (Bankier, 1986;Skinner, 1991).Applying the latter, one obtains a new estimator, usually called raking ratio (Skinner, 1991), which has the form where Skinner and Rao (1996) use a pseudo maximum likelihood approach to extend to complex designs the maximum likelihood estimator proposed by Fuller and Burmeister (1972) only for simple random sampling without replacement.The resulting estimator is given by where NPML ab (γ) is the smallest of the roots of the quadratic equation It is also shown that the following value for γ minimizes the variance of ŶPML .One can use the delta method to obtain a consistent estimator of the variance of this estimator of the form More recently, Rao and Wu (2010) proposed a pseudo empirical likelihood estimator for the population mean based on poststratified samples.Such estimator is computed as where, in this case, Ŷa abk , pB abk and pbk the weights resulting from maximizing the pseudo empirical likelihood procedure under a set of constraints (see Rao and Wu 2010 for details).Furthermore, η ∈ (0, 1).In this case, it is assumed that N A , N B and N ab are known, but this is not always the case.Authors also provide modifications to be carried out in (12) to adapt it to situations where only N A and N B are known or where none of N A , N B or N ab are known.In addition, auxiliary information coming from either one or both frames can be incorporated to the estimation process to improve the accuracy of the estimates.In addition, instead of an analytic form for the variance of this estimator, Rao and Wu (2010) propose to compute confidence intervals using the bi-section method described by Wu (2005) for one single frame and extending it to the dual frame case.This method constructs intervals of the form {θ|r ns (θ) < χ 2 1 (α)}, where χ 2 1 (α) is the 1 − α quantile from a χ 2 distribution with one degree of freedom and r ns (θ) represents the so called pseudo empirical log likelihood ratio statistic, which can be obtained as a difference of two pseudo empirical likelihood functions.
The R Journal Vol.XX/YY, AAAA 20ZZ ISSN 2073-4859   2013) extended calibration procedures to estimation from dual frame sampling assuming that some kind of auxiliary information is available.For example, assuming there are p auxiliary variables, x k (x 1k , . . ., x pk ) is the value taken by such auxiliary variables on unit k.Each auxiliary variable may be available only for units in frame A, only for units in frame B or for units in the whole population.In addition, it is assumed that the vector of population totals of the auxiliary variables, t x = ∑ k∈U x k is also known.In this context, the dual frame calibration estimator can be defined as follows where weights Then, with a similar approach to that of ŶBKA , another calibration estimator can be computed as x k = t x , being dk the weights defined in ( 5).
An estimator of the variance of any calibration estimator can be obtained using Deville's method (Deville, 1993) through the following expression where d k is given by (5) or by ( 14) according to whether we use ŶCALSF or ŶCALDF , respectively.In and e k are the residuals of the generalized regression of y on x.Some of the estimators described above are particular types of calibration estimators.For example, estimator (8) can be obtained as a particular case of ŶCALSF in the case where frame sizes N A and N B are known and the "raking" method is selected for calibration.Having noted this, one can use ( 16) to calculate an estimator of variance of (8).See Ranalli et al. (2013) for more details.
Table 1 shows a summary of the previous dual frame estimators according to the auxiliary information required.It can be noted that Hartley, FB and BKA estimators can be computed even when no information is available, but they cannot incorporate some auxiliary information when available.PML and SFRR can incorporate information on N A and N B , but PEL and CAL type estimators are the most flexible in that they can incorporate any kind of auxiliary information available.

Jackknife variance estimation
Variance estimation methods exposed so far depend on each specific estimator, so comparisons between variance estimations may lead to incorrect conclusions.Instead, one can consider jackknife, originally proposed by Quenouille (1949Quenouille ( , 1956) (see Wolter 2007 for a detailed description of this method in survey sampling) and extended to dual frame surveys by Lohr and Rao (2000), which can be used to estimate variances irrespective of the type of estimator allowing us to compare estimated efficiency for different estimators.
For a non stratified design in each frame, the jackknife estimator of the variance for any of the estimators described, generically denoted by Ŷc , is given by .Jackknife may present an important bias when designs are without replacement.One could, then, incorporate an approximate finite-population correction to estimation to achieve unbiasedness.For example, assuming that a finite-population correction is needed in frame A, a modified jackknife estimator of variance, v * J ( Ŷc ), can be calculated by replacing ŶA Consider now a stratified design in each frame, where frame A is divided into H strata and frame B is divided into L strata.From stratum h of frame A, a sample of n Ah units from the N Ah population units in the stratum is drawn.Similarly, in stratum l of frame B, one selects n Bl units from the N Bl composing the stratum.The jackknife estimator of the variance can be defined, then, as follows where ŶA c (hi) is the value taken by Ŷc after dropping unit i of stratum h from sample s Ah and Y Ah c is the mean of values ŶA c (hi).ŶB c (lj) and Y Bl c can be defined in a similar way.Again, one can include an approximate finite-population correction in any stratum needing it.In case of a non stratified design in one frame and a stratified design in the other one, previous methods can be combined to obtain the corresponding jackknife estimator of the variance.
Stratified cluster sampling is very common in practice.Now we illustrate the jackknife estimator when a stratified sample of clusters is selected.Suppose that frame A has H strata and stratum h has N Ah observation units and N Ah primary sampling units (clusters), of which n Ah are sampled.Frame B has L strata, and stratum l has N Bl observation units and N Bh primary sampling units, of which n Bl are sampled.
To define the jackknife estimator of the variance, let Y A c (hj) be the estimator of the same form as Ŷc when the observations of sample primary sampling unit j of stratum h from sample in frame A are omitted.Similarly, Y B c (lk) is of the same form as Ŷc when the observations of sample primary sampling unit k of stratum l from sample in frame B are omitted.The jackknife variance estimator is then given by: where

The R package Frames2
Frames2 is a new R package for point and interval estimation from dual frame sampling.It consists of eight main functions (Hartley, FB, BKA, SFRR, PML, PEL, CalSF and CalDF), each of them implementing one of the estimators described in the previous sections.The package also includes an additional function called Compare which provides a summary with all possible estimators that can be computed from the information provided as input.Moreover, six extra functions implementing auxiliary operations, like computation of Horvitz-Thompson estimators or of the covariance between two Horvitz-Thompson estimators, have also been included in the package to achieve a more understandable code.Finally, the package includes eight more functions, one for each estimator, for the calculation of confidence intervals based on the jackknife variance estimator.
A remarkable characteristic of these functions is the strong argument check.Functions check general aspects as the presence of NA or NaN values in its arguments, the number of main variables considered in the frames (that should match), the length of the arguments in each frame (that should also match) or the values for arguments indicating the domain each unit belongs to (which only can be "a" or "ab" for frame A or "b" or "ba" for frame B).If any issue is encountered, the function displays an error message indicating what the problem is and what argument causes it, so that the user can manage errors easily.Furthermore, each function has additional checks depending on its specific characteristics or arguments.The main aim of this exhaustive check is to guarantee validity of the arguments, so one can avoid, to the extent possible, issues during computation.
Much attention has also been devoted to computational efficiency.Frequently, populations in a survey are extremely large or it is needed to keep sampling error below a certain value.As a consequence, one needs to consider large sample sizes, often in the order of tens of thousands sampling units.In these situations, computational efficiency of functions is essential, particularly when several variables are considered.Otherwise, user can face high runtimes and heavy computational loads.In this sense, functions of Frames2 are developed according to strict efficiency measures, using the power of R to use matrix calculation to avoid loops and increase the computational efficiency.
Table 2 shows user and system times necessary to compute estimators using an Intel(R) Core(TM) i7-3770 at 3.40 GHz when different sample sizes are considered.Elapsed time is also included to get an idea about the real time user needs to get estimations.Functions of Frames2 have been implemented from an user-oriented perspective to increase usability.In this sense, most input parameters (which are the communication channel between the user and the function) are divided into two groups, depending on the frame they come from.This is to adapt functions as much as possible to the usual estimation procedure, in which the first step is to draw two independent samples, one from each frame.On the other hand, estimation details are managed internally by functions so that they are not visible for the user, who does not need to manage them.
Construction of functions has been carried out so that they perform properly in as many situations as possible.As noted in the introductory section, one can face several situations when using two sampling frames depending on their relative positions.Although the most common is the one depicted in Figure 1, cases shown in Figures 2 and 3 may arise as well.All estimators described but PEL can The R Journal Vol.XX/YY, AAAA 20ZZ ISSN 2073-4859 be modified to cover these three situations, so corresponding functions of Frames2 include necessary changes to produce estimates irrespective of the situation.
On the other hand, it is usual, when conducting a survey, to collect information on many variables of interest.To adapt to such situations, all functions are programmed to produce estimates when there are more than one variable of interest with only one call.To this end, parameters containing information about main variables observed in each frame can be either vectors, when only one variable is considered or matrices or data frames, when there are several variables under study.Cases in which the main aim of the survey is the estimation of population means or proportions are also very frequent.Hence, from the estimation of the population total for a variable, functions compute estimation of the mean as Ŷ = Ŷ/ N. To obtain the estimation of the population size, functions internally apply the estimation procedure at issue to indicator vectors 1 A and 1 B of sizes n A and n B , respectively.
To get maximum flexibility, functions have been programmed to calculate estimates in cases in which the user disposes of first and second order inclusion probabilities and in those other in which only first order ones are available, indistinctly.Knowledge of both first and second order inclusion probabilities is a strong assumption that does not always occur in practice.However, when calculating most of the estimators described in previous sections, second order inclusion probabilities are needed in many steps of the estimation procedure, mainly in computing estimated variances of a Horvitz-Thompson estimator or estimated covariances between two Horvitz-Thompson estimators.As an alternative, one can obtain variance estimations from only first order inclusion probabilities applying Deville's method reported in ( 16), by substituting residuals e k with the values of the variable of interest, y k .Covariance estimations are also obtained from variances through the following expression To cover both cases, the user has the possibility to consider different data structures for parameters relating to inclusion probabilities.So, if both first and second order inclusion probabilities are available, these parameters will be square matrices, whereas if only first order inclusion probabilities are known, these arguments will be vectors.The only restriction here is that the type of both should match.
As can be deduced from previous sections, an essential aspect when computing estimates in dual frames is to know the domain each unit belongs to.Character vectors domains_A and domains_B are used for this purpose.The former can take values "a" or "ab", while the latter can take values "b" or "ba".Any other value will be considered as incorrect.

Data description
To illustrate how functions operate, we use the data sets DatA and DatB, both included in the package.DatA contains information about n A = 105 households selected through a stratified sampling design from the N A = 1735 households composing frame A. More specifically, frame A has been divided into 6 strata of sizes N hA = (727, 375, 113, 186, 115, 219) from which simple random without replacement samples of sizes n hA = (15, 20, 15, 20, 15, 20) have been drawn.On the other hand, a simple random without replacement sample of n B = 135 households has been selected from the N B = 1191 households in frame B. The size of the overlap domain for this case is N ab = 601.This situation is depicted in Figure 4.
Both data sets contain information about the same variables.To better understand their structure, we report the first three rows of DatA: Each data set incorporates information about three main variables: Feeding, Clothing and Leisure.Additionally, there are two auxiliary variables for the units in frame A (Income and Taxes) and another two variables for units in frame B (Metres2 and Size).Corresponding totals for these auxiliary variables are assumed known in the entire frame and they are T A Inc = 4300260, T A Tax = 215577, T B M2 = 176553 and T B Size = 3529.Finally, a variable indicating the domain each unit belongs to and two variables showing the first order inclusion probabilities for each frame complete the data sets.
Numerical square matrices PiklA and PiklB, with dimensions n A = 105 and n B = 135, are also used as probability inclusion matrices.These matrices contains second order inclusion probabilities The R Journal Vol.XX/YY, AAAA 20ZZ ISSN 2073-4859 and first order inclusion probabilities as diagonal elements.To check the appearance of these matrices the first submatrix of order 6 of PiklA is shown.

No auxiliary information
When there is no further information than the one on the variables of interest, one can calculate some of the estimators described in previous sections (as, for example, (1) or ( 3 As result, an object of class 'EstimatorDF' is returned, showing, by default, estimations for the population total and mean for the 2 considered variables.In general, m columns will be displayed, one for each of the m variables estimated.Further information about the estimation process (as variance estimations or values of parameters involved in estimation) can be displayed by using function summary.
> summary (Hartley(yA, yB, PiklA, PiklB,  The previous output shows in the component Estimation the estimations of the population total and the population mean computed using the Harley estimator, that is, ŶH and ŶH .Estimated variances of these estimations, V( ŶH ) and V( ŶH ), are displayed in component Variance Estimation.
In the section Total Domain Estimations we can see estimations Ŷa , ŶA ab , Ŷb and ŶB ab .Estimates for the population mean for each domain, Ŷa , ŶA ab , Ŷb and ŶB ab are displayed in the component Mean Domain Estimations.Finally, θ, the estimated value of the parameter involved in computation of the Hartley estimator is shown.This additional information depends on the way each estimator is formulated.Thus, for example, extra information will include a parameter component when applied to a call to the Fuller-Burmeister estimator (and values of estimates for β 1 and β 2 will be displayed there), but not when applied to a The R Journal Vol.XX/YY, AAAA 20ZZ ISSN 2073-4859 call to the Bankier-Kalton-Anderson estimator (because no parameters are used when computing this estimator).
Results slightly change when a confidence interval is required.In that case, the user has to indicate the confidence level desired for the interval through argument conf_level (default is NULL) and add it to the list of input parameters.The function calculates, then, a confidence interval based on the pivotal method.This method yields a confidence interval as follows: Ŷ ± z α/2 V( Ŷ) where z α/2 is the critical value of a standard normal distribution.Only for the case of PEL, confidence intervals are based on a χ 2 distribution and the bi-section method (Rao and Wu, 2010).In this case, default output will show 6 rows for each variable, lower and upper boundaries for confidence intervals are displayed together with estimates.So, one can obtain a 95% confidence interval for estimations in the last two of the previous four cases in this way.For estimators constructed as (6), numeric vectors pik_ab_B and pik_ba_A of lengths n A and n B should be added as arguments.While pik_ab_B represents first order inclusion probabilities according to sampling design in frame B for units belonging to the overlap domain selected in the sample drawn from frame A, pik_ba_A contains first order inclusion probabilities according to the sampling design in frame A for units belonging to the overlap domain selected in sample drawn from frame B.
> yA <-with(DatA, data.frame(Feed,Clo, Lei)) > yB <-with(DatB, data.frame(Feed,Clo, Lei)) > > ## Bankier-Kalton-Anderson estimation and a 95% confidence > ## interval for the three main variables > BKA(yA, yB, DatA$ProbA, DatB$ProbB, DatA$ProbB, DatB$ProbA, DatA$Domain, + DatB$Domain, 0.95) Note that these examples include just a few of the estimators that can be used when no auxiliary information is known.As noted in Table 1, other estimators, as those in ( 12), ( 13) or (15), can be also calculated in this case.In this context, function Compare is quite useful, since it returns all possible estimators that can be computed according to the information provided as input.

Auxiliary information about frame sizes
For estimators requiring frame sizes known, as ( 8) or ( 9), it is needed to incorporate two additional input arguments, N_A and N_B.There is also a group of estimators, including ( 12) and ( 15), that, even while being able to provide estimations without the need of auxiliary information, can use frame sizes to improve their precision.The following examples show the performance of these estimators.
> ## SFRR estimator and CalSF estimator with frame sizes as auxiliary information > ## using method "raking" for the calibration for the three main variables As highlighted previously, both results match.Note that the argument met of the SF calibration estimator indicates the method used in the calibration procedure.The possibility of choosing the calibration method is given by the fact that computation of both SF and DF calibration estimators is based on the function calib from package sampling (Tillé and Matei, 2012), which can manage three different calibration methods, each one associated with one particular distance measure.These methods are: linear, raking and logit.
The requirement of knowing probabilities of inclusion in both frames for the units in the overlap domain may be restrictive in some cases.As an alternative, in cases where frame sizes are known but this condition is not met, it is possible to calculate dual frame estimators as ( 9), ( 12) or ( 13).Next, it is illustrated how to obtain some of these estimators with Frames2.
> ## Estimates for the three main variables using PML, PEL and CalDF > ## with frame sizes as auxiliary information in PEL and CalDF > PML(yA, yB, PiklA, PiklB, DatA$Domain, DatB$Domain, N_A = 1735, N_B = 1191) To calculate the PEL estimator, computational algorithms for the pseudo empirical likelihood method for the analysis of complex survey data presented by Wu (2005) are used.

Auxiliary information about domain sizes
In addition to the frame sizes, in some cases, it is possible to know the size of the overlap domain, N ab .Generally, this considerably improves the precision of the estimates.This situation has been taken into account when constructing functions implementing estimators (12), ( 13) and ( 15), so the user can incorporate this information through parameter N_ab, as shown below > ## Estimates for the three main variables using PEL estimator > ## with frame sizes and overlap domain size as auxiliary information > PEL (yA, yB, PiklA, PiklB,  Note that, in this case, calibration estimators provide the same results irrespective of the distance function employed.This is an interesting property that calibration estimators show only in the case in which all the domain sizes are known and used for calibration (see Deville 1993).

Auxiliary information about additional variables
On the other hand, some of the estimators are defined such that they can incorporate auxiliary information to the estimation process.This is the case of estimators ( 12), ( 13) and ( 15).Functions implementing them are also able to manage auxiliary information.To achieve maximum flexibility, functions implementing estimators (12), ( 13) and ( 15) are prepared to deal with auxiliary information when it is available only in frame A, only in frame B or in both frames.For instance, auxiliary information collected from frame A should be incorporated to functions through three arguments: xsAFrameA and xsBFrameA, numeric vectors, matrices or data frames (depending on the number of auxiliary variables in the frame); and XA, a numeric value or vector of length indicating population totals for the auxiliary variables considered in frame A. Similarly, auxiliary information in frame B is incorporated to each function through arguments xsAFrameB, xsBFrameB and XB.If auxiliary information is available in the whole population, it must be indicated through parameters xsT and X.

Interval estimation based on jackknife variance estimation
Finally, eight additional functions have been included, each of them calculating confidence intervals based on jackknife variance estimation for each estimator.To carry out variance estimation using the jackknife method, in addition to parameters to calculate each specific estimator, the user has to indicate through arguments sdA and sdB the sampling design applied in each frame.Possible values are "srs" (simple random sampling without replacement), "str" (stratified sampling), "pps" (probabilities proportional to size sampling), "clu" (cluster sampling) or "strclu" (stratified cluster sampling).Default is "srs" for both frames.If a stratified or a cluster sampling has been carried out in one of the frames, it is needed to include information about the strata or the clusters.Furthermore, the user is able to include a finite population correction factor in each frame by turning to TRUE the parameters fcpA and fcpB, set by default to FALSE.Since the main purpose of the functions is to obtain confidence intervals, parameter conf_level is now mandatory.As noted, these functions can be used, for example, to make comparisons between efficiency of estimators, as shown in the next example.
> ## Confidence intervals through jackknife for the three main variables > ## for estimators defined under the so called single frame approach with > ## a stratified random sampling in frame A and a simple random sampling > ## without replacement in frame B. The data set includes information about 7 variables: Drawnby, which takes value 1 if the unit comes from the landline sample and value 2 if it comes from the cell phone sample; Stratum, which indicates the stratum each unit belongs to (for individuals in the cell phone frame, value of this variable is NA); Opinion the response to the question: "Do you think that immigrants currently living in Andalusia are quite a lot?" with value 1 representing "yes" and value 0 representing "no"; Landline and Cell, which record whether the unit possess a landline or a cell phone, respectively.First order inclusion probabilities are also included in the data set.
> data(Dat) > head (Dat,3) Drawnby Stratum Opinion Landline Cell ProbLandline ProbCell 1 1 2 0 1 1 0.000673623 8.49e-05 2 1 5 1 1 1 0.002193297 5.86e-05 3 1 1 0 1 1 0.001831489 7.81e-05 From the data of this survey we wish to estimate the number of people in Andalusia thinking that immigrants currently living in this region are quite a lot.In order to use functions of package Frames2, we need to split this dataset.The variables we will use to do this are Drawnby and Landline and Cell.
> attach(Dat) > ## We can split the original dataset in four new different > ## datasets, each one corresponding to one domain.

Summary
The statistical literature about dual frame surveys started around 1960 and its development has evolved very quickly because these surveys are largely used by statistical agencies and private organizations to decrease sampling costs and to reduce frame undercoverage errors that could occur with the use of a single sampling frame.
Dual frame surveys can be more complicated to design and more complicated to analyze than those that use one frame only.There are several estimators of the population total available in the statistical literature.These estimators rely on weight adjustments to compensate the multiplicity of the units in the overlap domain.Some of these estimators allow to handle different types of auxiliary information at different levels.Nevertheless, none of the existing statistical software implements all of these estimators.
In this article we illustrate Frames2, a new R package for point and interval estimation in dual frame context.Functions composing the package implement the most important estimators in the literature for population totals and means.We include two procedures (Pseudo-Empirical-Likelihood approach and calibration approach) to incorporate auxiliary information about frame sizes and also about one or several auxiliary variables in one or two frames.Post-stratification, raking ratio or regression estimation are all encompassed as particular cases of these estimation procedures.Additional functions for confidence interval estimation based on the jackknife variance estimation have been included as well.
The functionalities of the package Frames2 have been illustrated using several data sets DatA, DatB and Dat (included in the package) corresponding to different complex surveys.We envision future additions to the package that will allow for extensions to more than two frames.
Finally, we would like to direct the reader to the package vignettes named "estimation" (Estimation in a dual frame context) and "formatting.data"(Splitting and formatting data in a dual frame context) for further examples and background information.

Figure 1 :
Figure 1: Two frames with overlapping.Figure2: Frame B is included in frame A.

Figure 2 :
Figure 1: Two frames with overlapping.Figure2: Frame B is included in frame A.

Figure 3 :
Figure 3: Frame A and frame B exactly match.
Inclusion probabilities are known in overlap domain ab for both frames.
with weights d CALSF k verifying that min ∑ k∈s G(d CALSF k , dk ) subject to ∑ k∈s d CALSF k

Figure 4 :
Figure 4: Frame and domain sizes for the data sets.

Table 1 :
Estimator's capabilities versus auxiliary information availabilityRecently,Ranalli et al. ( with ŶA c (i) the value of estimator Ŷc after dropping unit i from s A and Y

Table 2 :
User, system and elapsed times (in seconds) for estimators considering different sample sizes.
The R Journal Vol.XX/YY, AAAA DatA$Domain, DatB$Domain)) Using appropriate indicator variables as variables of interest, one can also estimate the overlap domain size, as shown below: and modified, given the number of cell phone numbers per individual.At the time of data collection, frame sizes of land and cell phones were 4,982,920 and 5,707,655, respectively, and the total population size was 6,350,916. computed