Chapter i SAMPLING 7.1 INTRODUCTION In discussing simple random sampling and stratified sampling desigrs - previous chapters we have seen that both these designs require somewz very laborious work in the sample selection Process. For both the des: an adequate sampling frame must be available and the sampling uns thereof must be numbered or otherwise identified so that a randomizac device, such as a random number table, can be used for selection of te J Specific_units. In such instances, a more simplified and versatile form 7 Probability sampling design, known as systematic sampling design, = be employed. If employed with care, the systematic sampling des¢ simplifies much of the work involved jin simple random sampling * stratified sampling. In addition to its being Operationally more convenie% than simple random Sampling, it ensures for each unit equal probability * inclusion in the sample. The basic idea of Systematic sampling is as follows: Suppose a sample “units is to be selected from a population of N units. Let these units © numbered from 1 to N in Some order. Let N=nk, where k is an integ called sampling interval. To Select a sample of n units, choose a unit random from the first k units and every k-th unit thereafter. Thus if a ut randomly selected happens to be numbered r, and the predeterminSee eee rere eee rccccet tt ae SYSTEMATIC SAMPLING 275 ampli interval is k, the sample will consist of units bearing numbers /, Te ts, tum an example, suppose that a population a > of 15 elements, numbered serially from 01 to 15 and that a random sample of 3 units is desired. To achieve this, select at random one af the first five units, 01 to 05, and then every 5" unit in the sequence. For \ example, if the first unit is 03, then the sample will consist of units 03, 08, and 13; if the first unit is 01, the sample will consist of units 01, 06, and 11. In this example, the sampling interval k is 5 and the sample size n is 3. The resulting sample is called the every k-th or 1-in-k systematic sample. Note that the selection of the first unit from among the first k units determines the whole sample. The procedure is termed linear systematic sampling so long as N=nk. If Nis not an exact multiple of k ie. N #nk, a systematic sample will contain either n or n-1 units depending on the serial number of the first selected unit, Such a sample will be called a circular systematic sample. We will discuss these methods in turn. 7.2 HOW TO DRAW A SYSTEMATIC SAMPLE Although _simple random sampling and systematic sampling both provide useful alternatives to one another, the methods of selecting the sample are different. A simple random sample from a population is selected by using a fable of random numbers. In contrast, various methods are possible in systematic sampling. We will first examine the case of linear systematic sampling procedure. raf ineat systematic sampling This is the case when N = nk. How is k chosen in a given situation? If the population size N is known, we can determine an approximate sample size n for the survey and then choose k to achieve that sample size. Suppose that there are 15,000 units in the population from which a sample of size n=750 is to be drawn. We must then choose k to be 20 or less. For k=20, we will obtain exactly n=750 observations, while for k<20, the sample size will be greater than 750. In general, for a systematic sample of size_n from ‘@ population of size N, Kus @ less than or equal to NV /n (Le. ksNW-n) an set k to be any number he/she wishes depending of the sampling fraction needed in the whole sample. If the sample would constitute one-half of the in the above example, the sample would be The sampler thus Cc upon the proportion k were 2, for example, population. if k were 20, as Sampling Methods-19276 AN INTRODUCTION TO SAMPLING METHODS only 100/20 = 5% of the entire population, which in this case is 759, Insy a case, the sample is obtained by taking every 20-th unit in the Population” The accompanying table displays a general scheme of k-systemat samples each of size n when N=nk. The important point to be noteg in the scheme is that, for a given numbering of the units, we are in effe selecting one group (or cluster) of units with equal probability of Selection namely 1/k from the following k groups or clusters forming the entire population. The table also shows the mean values of each sample. TABLE 7.1 LINEAR SYSTEMATIC SAMPLING SCHEME Sample No. Sample observations Probability | Sample mean ' Vi Vist Yoke ee V(n-i yee Wk 7 Y2+ Ve+2 V2 + V(n-i)e2 Vk A Vir Vivi Vaksi VMneri Vk k Vie VK V3 Vink Vk —______ —_—— The tabular arrangement of the samples guides us to provide a definition of systematic sampling as under: \ pefinirion ' @ population consists of the units 1,.y>....,.%y arranged " some fixed order and if the i-th Possible sample from the population 7 defined to be a subset with the units y, , Vari Vokes Yves + j=12- 7 then the subset selected by any means will constitute the «th systemalé sample of size n. The mean of the /-th s: ‘stematic sample is obtained as follows:SYSTEMATIC SAMPLING 277 ni S/S. Dvn 5 = 0 yy = : n (7.1 ene mea? of the k possible sample means EG, is T a knelt Vis jhe j=0 nk [7.2] Ne=nk, ¥,..is unbi. > this shows that when 1» Yyy IS Unbiased for 7 we illustrate the result with N=4 and n=2. With these choices, the sampling aterval K is 2 (=4/2). Let the values corresponding to the 4 units in the population be denoted by y,,2,3and y, The two possible samples and their means are shown in the accompanying table. TABLE 7.2 SYSTEMATIC SAMPLES OF 2 UNITS FROM 4 UNITS Neen eee ee eee ee ee ee er eee ncn ene EE Random# Sample values Sample mean 1 yoy noo re vt =g ity) 2 Yaa Tr=srt yo) mm The expected value of the systematic sample means Pay IS EG Wn 11 1 _Wtyet st Ya onde FH si)=4]50r tyy)t sr Da "22 Nonlin ic samplin: i encounter a oral problem wien N is not expressible in the form nk In this case, the present sampling scheme will give rise to sample of “equal size and consequently the probability of selection of the units enaing unequal. The property of unbiasedness of the sample mean also “not be ensured, name Consider the case when the population consists a " | a Vr noYyy. A sample of four units is to be chosen, tenths “ously, the sampling interval is 41/4, which is not an integer.278 AN INTRODUCTION TO SAMPLING METHODS following table illustrates the drawing of the possible samples with tg. 2.75 ~3 TABLE 7.3 _ POSSIBLE SYSTEMATIC SAMPLES FOR N=11, n=4 ted cece ee Random #. Units in the sample _ Random #._Units in the samp! 1 Vir Var V7 V10 2 Yr Vs VeVi 3 Vi YorV9 Note that the sample size does not remain the same in all Samples Fy. instance, the probability of selecting first and second samples is 4/11, whi, for the third it is 3 11. Let us check the unbiasedness of the systema sample mean. The three sample means are jp Litt yr tM 5, =r ee TH and 5 22 4 ~ Hence +h +5. EQiy.)= 7A 27 where paority2t-*Vn 11 This shows that Y,. is not unbiased for Y. 1,25 simple modification If we make a simple modification in the computation, the estimator becomes unbiased (Murthy 1967: 135). This is achieved by dividing the Product of the sample total and k by the population size. That is we form the i-th sample mean as follows: 5, - 4 (Totalof the ith sample) (7.3) Thus 3 — a 1 "Yah by), To 02 tS tig tM) 3 ~ 1's + Yo + ¥y) The average of these three means is uM vy 3 =¥, imSYSTEMATIC. SAMPLING » Fe ulation size. The method is described b, 10 7.2: Suppose that a 1-in- in 53 quires prior knowledge of the elow. k systematic sample is to be selected ‘sume that k is not an integer. Instead of taking 2 © ween 1 and k, we take a random numberr petween 1 and N. We then divider by k and determine the remainder. For instance, if the random number were 5 and k is 3, the remainder is 2 and “e begin with the second unit ¥2 If the remainder is 1, we begin with the first unit When the remainder is zero, we begin with k. We illustrate the concept with N=11 and k=3. The accompanying table shows the random number chosen, the ratio r/k , the remainder, and the sample values TABLE 7.4 MODIFIED SYSTEMATIC SAMPLE |. Random # rlk Remainder _ Sample values 1 3 1 Vi VaI7 V0 2 2/3 2 V2 V5 VeVi 3 3/3 0 V3 YoYo 4 4/3 1 Vy V4 V7-Yi0 5 5/3 2 Yr Vs Very 6 6/3 0 V3: Yoo Z 7/3 1 8 8/3 2 9 973 0 10 10/3 1 2 1 11/3 , Lu ich The Sampling distribution of remainders in the oe a soe imate, sample does not have the same probability of being sel ee irring, since the remainders 1 and 2 each has a probability of 4/11 o! a ) "ys Yio ach of them occurs 4 times resulting in the ae ee ae 2nd y2,.¥5,.%4,y4, Fespectively. The remainder 7er0 aces an being Units y,,y,,yy and consequently it has a probability as in the Selected, ‘We summarize the distribution of the rel accompanying table: :280 AN INTRODUCTION TO SAMPLING METHODS TABLE 7.5 DISTRIBUTION OF REMAINDERS Sample means Probability Sample values 4/11 Vi Va IT Vt0 =< _WAVat¥7 + V0 Vee 4 = Yt Vs tVetMu ya 4 V3 Vor V9 = {Wt Votv a 4 a SSF The expected value of the systematic sampling mean under the modifieg scheme is obtained from the above table as follows: Y2V5 VK vay 3/11 42 J E(¥ox) = +2 t+ Vow) Tt Af ntyaty t+ No i Yat s+ VetVu eS 11 4 I 4 1 a | a Nit yn tet LF 4 This shows that under the modified system, the sample mean is an unbiased estimator of the population mean when k is not an integer. This result will also hold true for total and the proportion. 7,94 Circular Systematic Sampling The procedure just described under non-linear sampling in section 7.22 has the disadvantage that it results in unequal sample size and consequently unequal probability of selection of the sample. As a result, the sample mean fails to become unbiased. A convenient solution to this - problem is to use circular systematic sampling procedure, which provides both a constant sample size and an unbiased sample mean. The method was proposed by Lahiri (1952), which we elaborate now. Consider @ population of N units as arranged round a circle and let_k now be the integer nearest to Nin. The procedure consists in selecting a unit by 4 fandom number between 1 and N inclusive and taking k-th unit thereafter going round the circle until the desired 1 units have been chosen. In this method, r can be selected entirely at random from among the N units, and the selection process is continued beyond N by renumbering units numbered 1, 2, 3, etc. as N+1, N+2, N+3, etc. up to N+r—-1.SYSTEMATIC SAMPLING 281 jllustrate the selection We al exhibit the following patter ST, ands ye ae CIRCULAR SYSTEMATIC GayoLES FOR N=11, N=4 Random Unit chosen Random Oni chosen number number 1 Yi Ye Wig ‘ mB em 4G 2 Yo Ys Ve yy 8 Ye Mi vs Ms 3 ys Yo Vg Wy 2 Te 4 Ya 97 Vig Ya 10 -— FF 5 Ys Ye Vrs a Yin Ys Ys 6 YX Yo My Wy : poo eoeocss Note that the samples are of equal size now (i.e. n=4), thus ensuring equal probability (here 1/11) of selection of the sample. In general, in selecting n units from a population of size N units, there are in all N samples, each having 1/N as its probability of selection. in contrast to k samples in linear systematic sampling. Consequently N = 1 = EG) = Ly SS i=l confirming that, under this sampling strategy, the sample mean is an unbiased estimator of the population mean. EXAMPLE 7,374 population consists of 5 elements 1, 2,3, 4, and 5. Drawa sample of size 3 and check the unbiasedness of the sample mean. i i to th SOLUTION: Here k = 5/3 = 1,67 which can be approximated to the next higher integer 2. Adopting the circular systematic sampling technique, the 5 possible samples and their means are as shown in the accompanying lable The expected value of the sample means is 1647410484) ieee =3 3 which equals the population mean, showjng that the sample mean is UNblased under circular systematic sampling. /DUCTION TO SAMPLING METHODS 242 AN INTRO! rasan pes Sac rears pie values Som os 2 2.4.1 7/3 3 3.5.2 10/3 4 4.1.3 8/3 5 5.2.4 1/3 6 SS 7.3 SYSTEMATIC SAMPLING IN RELATION TO OTHER DESIGNS 73 Ee Systematic sampling strictly resembles some other sampling. We make a comparative study of these relationships below. SYSTEMATIC SAMPLING AND CLUSTER SAMPLING: It seems intuitively clear from our discussion in section 7.2.1 that systematic sampling is essentially cluster sampling which divides the population units into k groups, one of which with all its units, is selected with equal probability so long as N=nk. Each unit y,(=1,2,....,N) in the population belongs to one and only one cluster. The probability of selecting a cluster is 1/k, which is therefore the probability with which any member of the cluster is selected in the sample SYSTEMATIC SAMPLING AND SIMPLE RANDOM SAMPLING: It should be noted that a systematic sampling is not a simple random sampling. It differs from the simple random sampling in that it does not give all possible samples of size n an equal chance of selection. Once the sampling fraction is determined, random selection of the starting point determines the whole sample. In simple random sampling with N=4 and n=2, for example, the 6 possible samples are ( 1, ¥2), (¥.¥3)s Me) (203). (Vey), ON) and any two distinct pairs of units (y,,¥,) have the same probability (i.e. 1/6) of appearing in the sample. By contrast, the systematic sample compositions are (y,,y;)and (y>,y,) and the samples (y\.¥). (M24): (Ys), aNd (ys, ¥4) appearing in simple random sampling, all have zero probability of being included in the systematic sample. Note that, while ct ewe sap ee ana opt foe agian ae lyk = N/n possible systematic samples of n units e Same size SYSTEMATIC SAMPLING AND STRATIFIED SAMPLING: Systemati nnn SAMPLING: Systematic sample may ait, ar oak eae as he opon drawing of one unit from a Hie ae pha and so on ang Selection procedure: in eyeiamat ming. the Gifference euletut Ic sampling, the units occur at the sameSYSTEMATIC SAMPLING 283 osition in the stratum, while in. stratifi ve ied sampli mncion js made within each stratum to select the units. es Lee 7.4 ADVANTAGES AND DISADVANTAGES OF SYSTEMATIC SAMPLING se gystemallc sampling is perhaps the most widely known design because of is rexibility 0 execution and simplicity in use. We enumerate below a few avantages and disadvantages of this design ya. Advantage’ OPERATIONAL CONVENIENCE: Systematic sampling is easy to draw and often easier to execute in the field. If drawing is done in the office or survey headquarters, there may be substantial savings in time. FIELD CONTROL: From the point of view of control over fieldwork, ‘systematic sampling offers great advantages over simple random sampling : 3. LESS NON-SAMPLING ERROR: Because of the fact that systematic ‘sampling is simpler to perform, it is less subject to interviewer's error than are either simple random sample or stratified sample, even if the survey personnel are less efficient. 4 REDUCED COST: Systematic sampling permits concurrent listing and sampling. Consequently, this can greatly reduce cost. This is in contrast with the simple random sampling, which requires a sampling frame in advance. Of course, it is preferable to have a sampling frame before the sampling operation, since systematic sampling can then be used to obtain precisely the sample size required. ? GREATER EFFICIENCY: Systematic sampling produces a more even spread of the sample over the population than does simple random sampling. This usually leads to greater precision to the systematic sampling except when the list is randomly ordered "42 Disadvantages The systematic sample is less representative 1 EFFECT OF PERIODICITY: than the simple random sample in situations when there is hidden it of the Periodicity in the population oF 4 cyclical or periodic movemen| data ic trend in the population EFFECT oF 1 . When there is a monoton REND Wy from the smallest t0 the tlements, that 1s, the population list variesETHODS NTRODUCTION TO SAMPLING METH AN II vice versa, a systematic sample will Perform le9y t or : lest element 284 large well. ERING: A major limitation of systematic Sampling '8 thay EFFECT OF ORD! y i it the ordering of the list, th 3. tion is made about "the unless some ee of the sampled elements does not Provig ee ‘oaivrating the variability of the sampling distribution, a basis for PLICATIONS OF SYSTEMATIC SAMPLING i ing is a common sampling design jn 1. IN GALLUP POLL: Se Galop oi iS a popular example. In the aoe ae ‘he Hore poll begins its sampling process by listing rice siete and then systematically selecting the districts for a ae study of households. _ 2. IN QUALITY CONTROL: The systematic sampling is extensively vsed in manufacturing industries for statistical quality control of their pro ucts Here a sample is obtained by taking an item from current production stream at regular intervals. 3. IN AUDITING: In auditing accounts, the most natural way to sample alist of accounts to check compliance with accounting Procedures or to verify deposited amount, is to choose accounts systematically for verification. 4. _IN MARKET RESEARCH: Systematic sampling has its wide applications in market research. Every 20" departing customer (say) at a checkout Counter may be asked his or her Opinion on the test, color, or texture of a food item. 5. IN CROP ESTIMATION: In crop yield estimation, systematic sampling has bee 'n extensively used, In estimating the Proportion of diseased trees, foresters may systematically sample field plots. 6. IN HEALTH STUDIES: In recent ye extensively used systematic sa Nutrition, iodine deficiency diso aSSE€Ss the Prevalenc ars, UNICEF and WHO have been mpling method worldwide in child rders, and arsenicosis surveys to @ Of these diseases, 7.6 SAMPLE MEAN AND ITS VARIANCE \ Let y,, denote the j-th member of the ith 1 and i=1, 2. ok With these not fesults ang theorems when Nenk, Systematic sample, so that <1. tations, we Present below a few( CO SYSTEMATIC SAMPLING 285 e i-th sample is = mean of th pI denoted by ¥,,and is defined as yo §, = n 7.4] jhe population mean is 2S, Y= ! nk (7.5) the variance of the systematic sample mean is given by _ k VG y= EG, -FP=1 YG, i=l 2 ) [7.6] since ¥) = Ly, /n , the variance expression (7.6) can also be put in the following form: oie iP] VO) = ay DAB A == En Bel) 77 For the estimated total, the variance is 2 ~ oN 2 ( yi} 0 V(Y%,)=—| pe — le Oy) =| Yor 7.8] THEOREM 7.1: In systematic sampling, with interval k, the sample mean J, is an unbiased estimator of the population mean PROOF: In systematic sampling, the whole sample becomes fixed as soon 4s the first unit is selected. Since the probability of selection of the i-th systematic sample is 1/k, each of the k samples has a constant probability Ik of being selected. By definition, the mean of the i-th systematic sample is ” iss (7.9] 7 n “uMmINg Over j and dividing throughout by k y, = 1 [7.10]AN INTRODUCTION TO SAMPLING METHODS 280 But k ae Ey) 745 Hence HGiy) 7 THEOREM 7.2: The variance of the mean of a systematic sample is N-l a2 k(n-I) go Mn)S Sat) 1g where 5 I kon _ 3 Sito) = Fqoy oy -5,) 713 s the variance among units that lie within the same systematic sample PROOF: By definition, the variance of ,, is k VGy)= EG, -FP “FG -7)? 17.14 Now consider the usual way of Partitioning the total sum of squares as follows. eit W-)S=Y Lo, -V)7=nDG, -¥Y + Y 0,-3.7 v1 j=l 7 1 7 Mk VF y)+K(=N) Sig) [From (7.13) and (7.14)] On transposing N- K(n=1) 9 W ay Slay) VPy)= Hence the proof. OROLLARY 7.1~°The variance expression (7.12) can also be expressed as follows. VOy) s? N 2 202 . =O oy) 7.48 where Is? 2 EO Siyy) NOn(y), the symbols bearing their usual meanings. The equation (7 15) implies that in order t ici i yr the within Sample variance must be eas efielent 7 Increased,v : SYSTEMATIC SAMPLING 287 E74 Suppose a population consists of 4 units 2, 5 40 1 Dy and 11.4 " ; ; eo sample of size 2 is to be di matic awn. Thi ase table is constructed with a few possi is implies that k=2. The wing a ble alternative ar 7 ine units the implied sample mean (7, ) and the variance of the mean o S. TABLE 7.7 SYSTEMATIC SAMPLE ESTIMATES UNDER ALTERNATIVE ARRANGEMENTS Units in the population _ Systematic sample . M ie 11, 10, 2,5 12 eg nance 10,5 75 5 2,5, 10, 11 2,10 6.0 1.00 5,14 8.0 : 5 2, 10, 5, 11 2,5 3.5 12.5 10,11 10.5 4 10, 2, 11, 5 10, 11 10.5 12.50 2,5 3.5 ae “Computed using (7.7) itis noted that with arrangement 1, the sampling variance is the least (0625). The arrangement 3 or 4 displays the most inefficient sampling resulting in the highest variance (16.25). This example is an illustration to demonstrate the importance of the arrangements of the sampling units in the population. The mean of a systematic sample is more precise than the THEOREM mean of a simple random sample if and only if Sy) >8? 61 PROOF: If 7 is the mean of a simple random sample of size n, V)= N-n S” O=W a while the variance of J,, from Theorem 7.2is . N k(n—)) 2 VPy)= Fy Sv(w) h ~ rf ‘us V(F,,.) < V(p) if and only if Nelo Ane Vgr Nom Ss” — Sin) = Orit N-n SS aTAN INTRODUCTION TO SAMPLING METHODS 288 Or if ( A(n =I) Saf) 7] _ a K(n=DSa(q) > Hl 1)? That is if Salo) 2S? This is an important result, which states that systematic sampling iS more precise than simple random sampling if the variance within the Systematy samples is /arger than the population variance as a whole. This can te achieved by rearranging the units in such a way that the units within Sect systematic sample are as heterogeneous as possible with TESPECt to the characteristics under study. What it points out that the arrangement - sampling units in systematic sampling is an important CONSideration - achieving a good estimate. This calls for a careful attention in Using systematic sampling. Example 7.4 is an attempt to demonstrate this feature. Further fS\aspect is examined in Example 7.5. THEOREM in alternative form of the variance of ¥,. in (7.6) is given by the following expression : 7 Se 2 "Gn)-Z f tab Z peony.) (7.17 where pis the intra-class correlation Coefficient between Pairs of-unts that are in the same systematic sample and its value depends on the arrangements of the units in the Population. It is defined as oe, ~F5 Pw a E(u, =F) 17.18 The numerator of t distinct pairs and t| denominator 18 (N lhe above expression is averaged over all N(n-1)72 lhe denominator Over all N values of y,. Since the YS iN, Py, Can be rewritten as 2 Py RENT ps7 Ly, -F\y, -Fy [7.19] "osu PROOF: By definition YGy)=TO SAMPLING METHODS AN INTRODUCTION 292 7 ¢ an +y)" Sp] Ld -7) rao FMI ~3,) pel ju n ai » jal 2 fle iste rd We DPS nk a DS +(n- Pos] = Sit Sax(¥o 7)b+er- Dena] n N =VGy)[-H(1-D uw] This proves the theorem. Note: If Py =O, VFy)=V Pu)» showing that systematic sample has the same precision as corresponding stratified sample with one unit per stratum: (7.25b] LATIVE PRECISION OF THE SYSTEMATIC SAMPLE MEAN Since V((J,,) 20, p, can not be less than —1/(n—-1). Thus the minimum value of p,, is -1/(n-1). in which case V(Jy,) =0. The relative precision (RP) of systematic sample mean with respect to simple random sample mean is given by ed a RP= 6 VOy) N= Fae Dp, 728 Clearly, the RP depends on the value of , w , in the following manner: (a) For p. ale ne Pw N-’ (y= (9), In which case the two methods are equally precise (b) For p< —! e hm N pO ) VG), in which case systematic sample is le 88 precise than Simple random sampl ple.SYSTEMATIC SAMPLING 293 . ision of s) i i vevative precision Stematic sampling method over the stratified Fag metnod Is Se au). 1 VAs YGy) [+ (n=Dp,, (7.27) ig evident from 7.25 (a) and 7.25 (b). s nee quation (7.27) has the following implications for varying values of p,, : yt Pe? 0, stratified sampling will yield more precise estimate of the pulation mean than systematic sampling wt Pw =0, both the methods are equally good. (MP < 0, systematic sampling will yield more precise estimate of the population mean than stratified sampling 7.5: Consider the following hypothetical population with N=30, ie5, and n=6. Each column represents a systematic sample and the rows ge the strata. TABLE 7.9 SYSTEMATIC SAMPLE DRAWN FROM A POPULATION OF N=30 Systematic sample numbers Strata Strata totals means Strata 1 2 3 4 5 y; > | 8 20 26 34 31 119 23.8 l 24 18 16 36 10 104 20.8 MW 6 19 26 21 23 95 19.0 Vv 19 13 12 8 35 87 174 v 25 10 a 4 28 131 26.2 M 7 8 29 7 23 74 148 laa go 68.136 ~=(147_—« 1506 10._—122.0 ¢ Mute the efficiency of the systematic sample mean with respect to the 80 ding simple random sample and stratified sample. % - . ted TION: The variance of the systematic sample mean y,, is obtained as ows, Sa > | Gy)=-v SSF SYSTEMATIC. SAMPLING rs to estimate the variance (7.6) “ wor avalible fo Us. But in our systomat ay eee oo ean Frys 'S simply one of the Several 5, , with the ae ee ile on which en number was chosen to start the sam ie ee e no information from our Sample concerning the canary of we which the systematic sample was taken, the elements with respect to the variable being measured. Lists arrai di alphabetical order may often be reasonably treated in this order, Under this condition, systematic sampling will be equivalent to a simple random sample of 7 units and logically, procedures developed for estimating the variance of estimates from simple random samples can be used. Under this assumption, the estimated variance of Vaux is represents a random ordering of N-n =——*;? N [7.29] WV ays where 2 n Xo, -5 Vyp-aP adel ~ zl n-l n-1 This estimator is biased but is satisfactory approximation to an unbiased estimator if the population elements are randomly ordered with respect to the characteristics under study. Sometimes the lists may be arranged in groups (i.e. students by grades, residents by socio-economic status) with variability expected between fS of survey variables. In this instance, the systematic Sample may peanalyzed as if it were a stratified sample. f Museum officials are interested in the total number of Persons who visit their museum during a 180-day period when an €xpensive antique collection is on display. Since monitoring the museum traffic each day is too costly, officials decide to obtain data every tenth day. The partial information from this 1-in- ic sample is ee n the accompanying table. Use these data to estimate the total number © Persons. visiti m during the specified period, Calculate the 98% Confidence interval for the population mean. —AN INTRODUCTION TO SAMPLING METHOD oz Day No. of persone visiting no. the museum 3 160 13 350 23 225 173, 290. SOLUTION Here N=nk=18%10=180, 1=18 and k=10. The other value, Sue are is is, Dy, = 1868, Yj = 1321450 jel fel An estimate of the population mean is eocee 270.44 where 2 adsl _ 1321450 -18(270.44) ou nl 18-1 Hence = 80-1 WP y) = : 8 (292.34) = 263.10 180 The standard error of the estimate is therefore ¥263.10 = 16.22 An estimate of the Population total is Y = NY, = 180(270.44) = 48679 20 The standard error of the total is therefore 180 < 16.22 = 2919.6. Hence 4 95% confidence interval of the Population total is 48679 20 | 2%OQ91ID6)<¥ < 4867920 +1 96 (2919.6) That Is 422987 sy x 54402 EXAMPLE 76. Ate deaths in 66 villages of MCH-FP area as registered under the KEDDRB's Health and Demographic Surveillance System in Matlab in 2000 are shown in the accompanying table.