p* ' $w&i& THE UNIVERSITY OF ILLINOIS LIBRARY *Mf NQN CIRCULATING CHECK FOR UNBOUND CIRCULATING COPY UNIVERSITY OF ILLINOIS Agricultural Experiment Station BULLETIN No. 153 COMPARISON OF METHODS OF SAMPLING CREAM FOR TESTING BY CARL E. LEE AND NELSON W. HEPBURN URBANA, ILLINOIS, FEBRUARY, 1912 SUMMARY OF BULLETIN No. 153 1. In testing cream, duplicates obtained by the same or different testers are accompanied by a reasonable variation. In samples reported 87.7 percent were exact duplicates or checked within one-half percent, 9.8 percent varied one percent from the individual, while 2^ percent showed a variation of more than one percent. The above figures are significant to anyone expecting exact duplicate results each time a sample is retested- Pages 550-551. 2. There is a greater latitude of variation between composites and indi- vidual samples than there is between two sets of composites, but this variation is approximately evenly divided. Pages 552-553, 554-555- 3. Of Hepburn's composites 36.40 percent tested above, while 41.99 percent fell below the individual. Pages 552-553. 4. Two series of composites taken in the, same manner showed the follow- ing tendency for variation : 63.73 percent were exact duplicates, or varied but one-half percent; 17.60 percent varied i percent; 6.33 percent varied 1.5 per- cent; 7.52 varied 2 percent; 2.1 percent varied 2.5 percent; while 2.8 percent varied more than 2.5 percent. Page 565. 5. During the winter months 41.87 percent of the composites tested lower than the individual against 35.38 percent testing higher. During the summer months 33.91 percent tested lower against 43.16 testing higher. Page 570. 6. The results of long time averages are seen in the figures showing that during Lee's period of six months only 15.7 percent of the samples varied more than one-half percent, that during Hepburn's period 15.9 percent of the samples varied more than one-half percent, and that under the proportionate system, for a period of six months, 24.75 percent of the samples varied more than one-half percent. Page 572. 7. Based on the year period only 5.95 percent of Lee -samples varied more than one-half percent, while during the same length of time 7.78 percent of Hepburn's varied more than one-half percent. Page 572. 8. Results from the yearly average butter fat show 14.52 percent of com- posites below the individual and 7.7 above, by Lee, and by Hepburn, 15.34 per- cent below and 7.78 above. These results are closer than the results obtained by duplicate testing. Page 572. 9. Results are corroborated by total pounds of butter fat in Table 16, showing a variation between individual and composite samples of 27 hun- dredths and 16 hundredths percent for Lee and Hepburn respectively. Page 574. s/TY OF ^'CULTURE !r > . COMPARISON OF METHODS OF SAMPLING CREAM FOR TESTING BY CARL/ E. L/EE, ASSISTANT CHIEF IN DAIRY MANUFACTURES,* AND NELSON W. HEPBURN, FIRST ASSISTANT IN DAIRY MANUFACTURES INTRODUCTION For several years past butter fat, as determined by the Babcock test, has been used as a paying basis for the dairy product deliv- ered to creameries ; yet, after years of experience, those who have watched the development of the various systems of testing and sampling are realizing that many of those systems approved or disapproved by common consent, may or may not be giving results which are mutually satisfactory to both manufacturer and patron. Since butter fat is the basis of all creamery calculation and, in turn, the basis of payment for the product from the farm, it is im- perative that the method, or methods, followed in taking samples and determining the percentage of fat be consistent to a degree which will work hardship neither to the manufacturing concern nor to the patron of the creamery. It is conceded that the Babcock test, properly manipulated, is the easiest and the most accurate and economical means in the hands of the creamerymen for testing milk and cream for fat, and that the points of dispute and subjects for popular and scientific discussion are more intimately associated with systems of sampling. It is not the purpose of this investigation to advocate any one method of sampling as being accurate above all others; for it is perfectly clear that if it should appear that any or all the systems give equally good results, they would not all be applicable to every condition. For example : our larger, and many of our smaller, centralizers pay for each delivery or shipment of cream. This, in itself, presupposes the testing of every can of cream as it arrives, and the system of sampling automatically regulates itself. It will further appear that the experience of the operator, the number of patrons, and the business principles followed, all modify the test- ing system. Now Assistant Professor of Dairying", University of Wisconsin 545 S7 546 BULLETIN 153 [February, IMPORTANCE OF TESTING There is a common opinion among creamery inspectors that test- ing does not occupy the prominent place it should in creamery practice, indicating that few are familiar with the disastrous re- sults of careless testing and practices of sampling which may lead to an inaccurate determination of butter fat. If the test applied to creamery weights is to be used as a basis for payment, then it bears the same relation to that industry that a system of weights and measures does to any commercial business. Few realize that the creamery overrun will, on a. 40 percent cream basis, be in- creased or decreased 3 percent by introducing an error of I per cent in testing. This means, in a plant averaging 700 pounds of butter daily, a loss or gain of 21 pounds of butter, worth about $6.00 the salary of two gQod men. Thus it becomes evident that any practice resulting in a test uniformly high is a source of constant drain to the plant. Should the error fall in the opposite direction, the plant must sooner or later suffer from a list of dis- satisfied patrons. As already referred to, the method of sampling cream for the determination of butter fat has been a subject of dispute and doubt, particularly since the change in manufacturing conditions has led to the delivery of a large amount of butter fat in the form of cream. The object, therefore, of this investigation is primarily to compare the efficiency of systems of sampling now in use, to test their reliability, and to form some opinion of their application and relation to creamery practice. METHODS OF COLLECTING DATA The data presented in this bulletin cover the work of one year in the creamery operated by the Department of Dairy Husbandry, University of Illinois. As a result of this method of gathering material several topics of secondary interest have presented them- selves, some of which seem to bear a direct relation to the subject in hand. Wherever possible such influences have actually been taken into account, but many of the facts that appear can be assigned no numerical value and at best one may only call attention to their possible influence. If the sole object of the investigation had been to make mere comparisons on the various systems of sampling cream, this prob- ably could have been more successfully and more scientifically car- ried out entirely distinct from the commercial activity; but with the secondary object in view it seemed best not only to compare the systems on a large scale, but also to correlate them as closely as possible with modern creamery practice. This end was accom- plished, as each sample represented not only a comparison but also a can of cream which contributed to the creamery's daily supply. COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 547 During the year the creamery had on its list some 103 patrons but the results published are on only 77, as the other 26 began de- livering toward the end of the season and would not make fair comparisons with patrons delivering cream the whole year. The output was, on an average, about 2500 pounds of butter per week. KINDS OF SAMPLES TAKEN The comparisons drawn in the following data are between "in- dividual" samples, that is, samples taken from each delivery and tested at once; "ordinary composite" samples; and "proportionate" samples, that is, samples taken in proportion to the amount of cream received. The composite system of sampling here referred to con- sists merely in obtaining a sample composed of representative amounts of cream from each delivery of a single patron for a period of fifteen days the method employed in whole milk creameries. The samples were kept in one-half pint lightning jars with tin covers, and preserved with corrosive sublimate, in tablet form, two small tablets being used during the hot months and one during the cooler season. In making the composite no effort was made to ob- tain a sample in proportion to the amount delivered; the object being merely to have a sample composed of cream from each de- livery at the end of fifteen days, as well as a sample large enough for convenient manipulation in testing. It may be added, how- ever, that this practice, when averaged, usually results in securing amounts of cream from each delivery which are approximately equal in volume. Since the primary object was to compare individual and com- posite sampling, the first step was to obtain a representative sample of cream from each patron's delivery. This was usually a half pint taken from a can of cream after it had been vigorously stirred to insure thoro mixing. These samples were then tested for fat, and composites made by adding approximately equal amounts of cream for each delivery to the respective composite bottles; that is, if patron Number One delivered a can of cream a half pint bottle was taken for a sample, tested, and about one-fourth of it added to composite bottle Number One. The method of handling the proportionate sample varied from this only in that an effort was made to secure samples in proportion to the amount of cream delivered. DETAILS OF TESTING For the whole experiment 40 percent 1 8-gram bottles, gradu- ated to one-half percent were used. Scales for weighing samples were of the creamery torsion type. The duplicates represent the results of two men working independently, rather than the close- 548 BULLETIN 153 [February. ness of duplication by one man. This offers the additional ad- vantage of making it possible to form some opinion of variations occasioned by two persons doing the same testing. In preparing the composites for testing they were handled in lots of 24 each, the capacity of the centrifuge. The samples were removed from the shelf and placed in rotation in a pan of sufficient height to permit the addition of hot water to the cream line. After warming to fluid condition the cream and fat were removed from the sides of the bottles by means of a spatula, after which a uni- form sample was obtained by pouring several times from one vessel to another. An 1 8-gram sample was then weighed into the test bottle, after which the composite was passed to the second tester, who obtained a duplicate. Subsequent to the sampling the testing, with few exceptions, was carried on in the usual manner. Highly satisfactory results were gained, however, by a slight change in manipulation previous to the whirling. This consisted of filling the test bottles with water to the base of the neck, previous to the first whirling. Experience in testing cream indicated that the use of such a method resulted in a larger proportion of clear tests, and the absence of those com- mon faults described as curdy samples and burned fat. It may be noted in this connection that the testing of cream, and especially of composites, during the summer months is frequently attended by undesirable results in the form of burned samples or otherwise cloudy tests. It will be noted that each system of sampling is practically a prob- lem of its own and will necessarily be treated as such preliminary to comparing it with any other system. In the general comparison the individual sample is used as* the basis for comparison. Since all the data are taken from the creamery records, the comparisons appear first in terms of pounds of butter fat. Such comparisons are interesting from the standpoint of the problem of financial loss or gain accruing to each patron, but they are not on a comparable basis, as the pounds of cream for each patron might be widely different. It thus becomes evident that a system of sampling occasioning a large number of pounds variation, in spe- cial instances might, or might not, be the cause for a high percent- age variation. In view of these facts it seemed best to make a preliminary study of each system, and then compare it with the in- dividual system on the percentage basis. Since each composite repre- sented the test of a patron's cream for a half month, the individual samples would be reduced to similar terms by dividing total pounds of butter fat for those fifteen days by the total pounds of cream, in other words, determining the "average test." The average test, representing the individual samples, is then fittingly compared with any composite taken for the same length of time. /p/ Cu V w O O\iO O CO 3S 001> O t^ O\ M VO 00 W 1C 00 ON t> iO TH CTl-rOrHrHrHM^CrOrHMM VO rH r<3 tron mber M M M M M M M M M M <*3 rO rO rO rO 1912] 553 rH rH f>) CO f>J ^tCOrHrH S M CJ CO Cl * f>) CO CO V/) rH Cl rH M CM M rH * 00 rH CO rH rHrH COrHM rHrHrH rHrOrH 8 554 BULLETIN 153 [February, O g M P H 8 o O S g a C/3 Q > t-i Q 2: 10 ? 13 * rH rH rH M rH jd rt .S CO rH rH rH rH N o & a S rHrHr- rH rH c o M rH rH . rH O C, rH rH rH 0) rH rHrH rH fNCOfMrHCO CO VO COt^l rH rHTj-rHMrHrH - ^M CO rH CM fMMrHrHCOfNrHVOCvJrHfNrH r3 rH rH M rHrH * rO*/3rH rH rH if) * t- M CO * * fN CO ^O CO N rH M * (S CO rH CO rH d CO rH 5* rfrH T* .*Tj-rHCOlOt>>'lO^OrHfX100'*OcOfMC'<}-fxlrHfN-*rl rH rnrH.0 rHrH^COCOdVOcOnrH^^nrHrH **** \M *c3 rH * ."2 N l/}-HrH rHrHfNrHrH rH rHrHrHrHrH ; S rH rH rH rH rH rH CO rH rH C 0) o d CO rH rH rH 0) fin - rH J rH C ,0 C re P. d 3 G 555 8 TH n ro TH Tt-C4f<3-*-TH- TH fC rH ,-1 TH TH TH TH TH fO rH * N M rHTHMTHC^rHrHTH TH rH tO rH 8 556 BULLETIN 153 [February, Referring to the summary of Table I it is seen that 87.7 percent of the 1165 samples reported tested the same or fell within the limit of one-half percent variation, while there is only one instance where the difference was as much as 3 percent. It is further a fact worthy of comment that the differences are approximately evenly dis- tributed on either side of the zero line, which means that in any series of such tests these irregularities tend to correct or at least to equalize each other. Table 2 is a distribution table used as a basis for comparing Hepburn's composite tests with average individual tests on each patron, for periods of fifteen days each during one year. The data are arranged similar to those in Table I. The zero column shows the number of composites testing the same as the average individual tests. Reading to the right of the zero column gives the number of samples with their respective percentages of fat above the individ- ual, while those below the zero column give the number of samples with their respective percentages of fat below the individual. In studying the comparison of composite with individual sam- ples it is at once obvious that a much wider distribution is displayed than in the comparison of duplicate testing, but again, there is a remarkable equilibrium of total results derived from the tendency for the same number of samples to vary an equal amount in either direction. Out of a total of 1066 samples used for comparison 36.46 per- cent tested higher than the individual, while 40.30 percent fell be- low the individual, and 57.77 percent either tested the same or came within the limit of one-half percent variation. Of the composites 19.22 percent varied one percent from the individual; 10.5 percent varied 1.5 percent; 5.62 percent varied two percent; while in 6.92 percent there was a variation of more than two percent. The high- est variation recorded in this comparison is in two samples differ- ing 4.5 percent from the individual. Special attention should be directed toward the figures showing that 40.30 percent of the composites tested lower than the individ- uals and only 36.46 percent tested above the individual. While in this investigation we cannot hope to assign a cause to such a tendency, its presence is noted in every comparison. Table 3 is a complement of Table 2 and needs no explanation. Since this is a comparison of Lee's composite with the individual samples, it should be noted that there is the same degree of cor- relation between Table 2 and Table 3 as between the comparison of testers in Table i. i<)i2\ COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 557 The variation of Lee's composite from the individual is prac- tically the same as in the comparison of Hepburn's composite with the individual. 55.7 percent of the samples tested the same or fell within the limit of one-half percent variation ; 20.3 percent varied one percent from the individual; 11.7 percent, 1.5 percent; 6 per- cent, 2 percent ; while 6.39 percent varied more than 2 percent. The highest variation was 5 percent in one sample. PROPORTIONATE SAMPLING The comparison of the individual samples with the proportionate composite is on a slightly different basis from those comparisons in- volving individual and ordinary composites, since the data covers a period of only six months and consequently represents about half the number of samples used in former comparisons. The proportionate sample was taken by means of a graduated pipette, the size of the samples varying in different instances from i cc. per pound of cream to 10 or 20 cc. The amount of cream usually delivered by the patron was the standard determining the ratio to be used, and the effort was made in this case, as with the regular composite, to have a sample at the end of fifteen days large enough to be convenient for manipulation in testing. In this we were successful only in so far as we were able to estimate the amount of cream to be delivered by a patron for fifteen days, and so it often happened that the sample for the testing period was represented by a smaller amount of cream than is suitable for the best results. Theoretically, composites taken in proportion to the amount of cream should give results corresponding most closely to the in- dividual testing, and such would probably be the case in sampling milk or thin cream, but the mechanical difficulties attending the use of such a system with heavy cream tend to destroy its proportion- ate value, and with thick, sour, and viscous cream, such a sample often becomes proportionate only in name. This perhaps will serve to explain the cause for some of the irregularities of the propor- tionate sample taken in this manner. Table 4 illustrates by distribution the difference between pro- portionate composites and individual samples. In this, as in for- mer cases, the individual is used as a standard for comparison. 558 BULLETIN 153 [February, fc hH W H $ W g S Ifl B u, ? ) 3 - rH rH rH 3 c to 0) o to M rH N rH rH (4 4J C s rHrHM rHrH rH rHrH /3 rHrHrH rHrf C rH - C-1 MrH MnrH rHrHrHrH rH rH .s S it CJ J3 *a3 ,0 to rH rH rH "E 0) o ^ (O V & Tf rH rH ? rH rH , rH c fc j R d C 1912} COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 559 rH 1-1 rH M * rH rH rH rH rH rH rH * rH VO M rH rH 1-1 M rH w w' rH T-I rH r-> rH rH 8 VO f1 N rH W iO x/} M O* rH ^J- M VO rH rH rH CO f<3 rH CO M M rH rHCOMrHrH MfOM CS C^rHMrHrHr^C^ClrH | rH o^ rH ON >0 i> * rH VO iO rH 'O * M rH rH rH rH t^ CO rH rH rH M rH CO if) rH ro W5 rH M >O^OI>OO^OrHMrO-HrlOOOOOrHf-lrr)-*VOOt^OO<^OrHPr)-*OO\OrHl^ rOCOf3t<5fOTtTtri-^-T}-'^-Tj-^t'ioiOLOioiOiOU^iOiOi/)^CO^O v vo i Ot > -l'-I> *c3 4-> O H Average percent 560 BULLETIN 153 [February, p w < < n O O 05 fe M P - 2 w 8 O M O fc -w ^ o M b ft O fc i M ft I U3 * rh H to \ * rH 3 & P. CO rH W e CO o .0 (4 iR M CS rH rH rH rH -*J a a; o M CO M rH M rH w (e ^ M rH rH Cl rH M ^0 "3 ,0 ro rH rH r-i -t-> a n> o ^J to rH rH V h * iR * VO ! 1 H H - OJ n rHr-iTt^coocriOrHcifo-*icoooNOrHrifO'tiooi>GOC^OrHr-iro^io ( P ^5 J c 1912] COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 561 MfMrHCrH rH tS)C rH D "\ > ^ oo oo vo ON VO CO CO s - 4-1 ^ n vo oo C ~ O \J g & t^ o n i-- rH ri rH * N CO* n co co rH ON O !>; 00 O rH co O *' ON" ON ON 1 5 * 00 rH CM rH O ON TJ- ^ "! . rH rH * rH rH O rH O $ 00 00 rH CO - rl - CM 00 * 00 rH X S" ON vo 00 rH- rH 5} rH 00 CO ^T rH uj O ri o ri rH rH rH r^ ^ io' * ON O * CO vo" oo vo* 1* N o ri CO vo CO *<3 \N M 4-) t^ iO rH rH rH ri ri fl s *> JH 00 CO O ON ON vo oo r) io rH rH n. \N *PS CO cu o. s cu a - u a, c -o ^4 W y M o .S 1 5 CX rt ^ n 1 11 JJ& a "S ' t! HH ' ^ii^&i-Si S.ti-a '.-s 2 S.tJ | a^.S^^a-Cs^o COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 563 Table 4 shows a still greater latitude of distribution : viz., 44.77 percent, against 55.7 percent in the previous comparison, tested the same or varied only one-half percent; 20.29 percent varied I percent; 14.31 percent, 1.5 percent; 9.16 percent, 2 per- cent; and 11.31 percent, more than 2 percent. The widest variation was 5 percent. Here again, with 44.12 percent of the samples test- ing below and 36.65 percent testing above, is a tendency for the composite to fall below the individual. No doubt the wider latitude of variation, as well as a greater tendency to collect on the low side in this system of sampling, is largely accounted for by the fact that in an effort to secure a proportionate sample where small amounts of cream were delivered, it often happened that small samples were collected for testing. This practice sometimes resulted in samples which were too small for convenient manipulation. The foregoing distribution table (Table 5) is prepared merely to show how the proportionate composite varied from Hepburn's regu- lar composite. Whether this comparison w^ould be any criterion for the comparative results of proportionate and ordinary composite samples in general, would depend largely on how the proportionate samples were taken. The notable feature of this table is the tend- ency of the proportionate composite to fall below the regular com- posite. Table 6 is the percentage summary of distribution tables num- bers i, 2, 3, 4 and 5, compiled here merely to show their compara- tive values. Figures in the table show percentage of samples vary- ing in either direction from the zero line by each system. 564 BULLETIN 153 [February, Table 7 is a comparative expression, in terms of percent, of the variation occasioned by the methods, with the individual used as a basis for comparison. Column i gives the percent of samples test- ing the same as, or varying one-half percent from the individual; column 2, the percent varying one percent; column 3, the percent varying 1.5 percent; column 4, the percent varying 2 per cent; col- umn 5, the percent varying more than 2 percent. Columns 6 and 7 are quite significant in that they show the total percent of samples testing above and below the individual and are consequently the one quantitative expression of the difference between the various composites and their relation to the individual. TABLE 7. SUMMARY TABLE COMPARING THE DIFFERENT SYSTEMS OF SAMPLING WITH THE INDIVIDUAL SYSTEM 1 2 3 4 5 6 7 Percent same or varying %% Percent varying \% Percent varying 1-%% Percent varying 2% Percent varying more than 2% Percent above individual Percent below individual Hepburn with individual 57.77 19.22 10.50 5.62 6.92 36.46 40.30 lyee with individual 55.70 20.30 11.70 6.00 6.39 34.30 41.99 Proport'n- ate with individual 44.77 20.29 14.31 9.16 11.31 36.65 44.12 1912} COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 565 DIFFERENCE IN SAMPLES TAKEN IN THE SAME MANNER From the foregoing data it is evident that there is considerable difference in tests resulting from the various commercial methods of collecting samples, which naturally leads up to the question of how much variation there might be between samples purported to be taken in the same manner. Such data would be especially in- teresting on the ordinary composite, since this is the usual cream- ery sample. It would be further suggested that any variation ob- tained under different systems of sampling should be in excess of this if it is to be attributed to the difference in method. As a means of obtaining some idea of the variation in samples taken in a like manner two ordinary composites were collected during the last month of the experiment. These were taken after the manner de- scribed for collecting the regular composite, so that either might have been used as a paying basis for the month. The results here showed 34.6 percent of the samples to be exact duplicates; 19.08 percent, one-half percent high; 9.86 percent, one percent high; 2. 1 1 percent, one and one-half percent high; 4.22 percent, two per- cent high; 1.4 percent, two and one-half percent high; and 1.4 percent, four percent high. 10.5 percent were one-half below the individual; 7.74 percent, one percent below; 4.22 percent, one and one-half percent below; 2.83 percent, two percent below; 0.7 per- cent, two and one-half percent below; and 1.4 percent, three and one-half percent below. By subtracting the variation resulting from duplicate testing, we then have some indication of the variation accruing to samples taken in the same manner. This, however, is perhaps only a sug- gestion of results, since the number of comparisons are too small to warrant any conclusions. SEASONAL INFLUENCE The condition of composites, together with some knowledge of the causes for such conditions, is probably responsible for the in- ference that seasons of the year, accompanied by various intensi- ties of heat, cold and light, have their effect upon the results ob- tained from the composite samples. As a means of comparing the effect of seasons two distribution tables, numbers 8 and 9, are here presented, comparing composite with individual samples, for the winter months of December, January, February and March, and the summer months of June, July, August and September. 566 BULLETIN 153 [February, S Q < u, ? a - 3 "O H iH CO rH rH rH T> .9 V CO rH rH rH > O & n) 5 rH rH rH 4-1 a V o N rH CM rH C rH rH 0) AH rH rHrH ,_(,_|,_| ,_( ^^fNlrO rH rH rH i-tNCl rH CM COrHrHrH * CM CMCNCOcNrH CM COcMrH CMrHCM CM rH CM CO 1O fMCMfM rHrHTj-Tj-rH CM CM rHrHrHCMCMVO CO rH CO CM \ MrHCO CMcOrHCMT-trH CMcOrHrH CMCqcM rHOJ tt CM CM rH rHCNrHrO CMCOrHCMrHrH rHCMrH "*CMrH 13 rH rH rH CMCMrHfM rHrH rHrH MH > H rH rH rH rH rH rH rH "C d rH fi s rH rH rH rH j3 "3 ,a CO rH "c V o CO 0) OH - rH rH 5 E h V 1 K p. a 1912} COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 567 rH rH rO rO rH rH M CO >-Hf<3rOfOrH 8 s MMi-HrH rH rHrHrH s 568 BULLETIN 153 [February, tf J 3 4 1 5 5 j 4 J 5 5 3 4 4 4 J 3 I | 3 ) 4 > 1 < ' 1 ; r i i i 1 i 1 4 4 i 4 4 1 1 ^ > 1 I i! 1 5 5 i } 3 4 < Percent above individual 10 $ H * - ro rH rH CO rH rH 5 rH rH rH M M r-INrH rHrH MrHrH rH drHrHrH rHrHrH rHrHrHtO rH rH rH MrHrHrHC<)rHrH M rHrH T-trHCrH rHrH rH rHrHfO O tOrHrHMMrH t^tSrHMdCl-* -*rH MMMMrHrHtOrH rH rHrH Percent below individual * rHMrH WrH rHrHWMrH rHrH N MtO VO rH rHrH rH *^r- ------ rH C^ M rH rH rH rH rH rH rH rH rH rH N rH rH rH to rH to - 5 - rH Patron number 1912} COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 569 rHCOrHCJrHOlrHrHf^ rH rt rfr f<3 N rH rH rH < * co CM 00 l> t^ co 3% i i rH * 00 CO CO rH N W vo * rH rH ri * a? rH %S rf VO' rH t^ CO vo * 00 M' rH oo ON CO ** VO '*' rH rH O iO I> in 10 M' CO CN M Percent below individual * 0> rH \T> i> ^6 rH rH rH 38 N t>; rH * rH f>l 00 C4 rH l> 1C CJ O O\ ^J- iO ri ci * M * o "^ . rH rH CO 10 N O\ iO * f~> ^J- fM "t 9 ^ Tj- iO Winter months Summer months a . c n3 00 V ^ O rH re 3 SH *O )#)-< O *U r> 00 VO CO rH a MH O - O *fl> | cc-| 00 O\ III rH co' * CO in *R^ III vo t> INFLUENCE ercent riation VO CO 11. SEASONAL rV N > percent iriation iO O 10 rH rH CJ w r^> rH rH n Same or ^ percent variation CO CO O vo vo' * months ;rmonths Winter Summ< 1912} COMPARISON METHODS OF SAMPLING CREAM FOR TESTING 571 Table n is also derived from Tables 8 and 9, illustrating still further the comparative relation between composite and individual samples collected during the two extreme seasons of the year. The study of seasonal influence is, in this connection, of pe- culiar interest, since the question of difference caused by season is intimately related to influences of evaporation and other factors potent in causing disagreement between composites and individu- als. Referring to Table n, we note a tendency for the composite to fall below the individual in winter, with more than a correspond- ing tendency for a higher test in summer. RESULTS OF AVERAGES Tables 12 and 13 are summary tables derived from data re- corded under Table 16, which gives total butter fat, cream, and average tests by patrons under each system of sampling for six months and for one year. It might be added that the object of in- serting the six-months comparison is to show data comparable with the proportionate samples, since they were collected only for that length of time. TABI.E 12. COMPARISON OF COMPOSITE WITH INDIVIDUAL TESTING FOR Six MONTHS 2^ 2 1#- 1 % X 1 IK 2 1Yz Number of variations according to L