a_i DESIGN OF SIMPLE BIOLOGICAL EXPERIMENTS AND ANALYSIS OF ‘VARIANCE 10 Introduction Experiments, are performed ann by researchers of investigators in Virtually all OF fields of study, usually to discover something about a particular Process oF system, ‘Gterally, am experiment is any ‘Process or action that leads to a Well defined result called ‘outcomes or responses. More formally, Sm be deine as a study in which a treatment, “Mntionlly introduced ito the input varales 2 Process or system, so that We iy observe and identi the reasons for changes tht may be obese he result or outcome (ourput variable). Inthe design ofa good experiment, there are ‘scaly some importan steps tonote whch nck 4 experiment procedure, ot program is, U0] Define the objectives. Record (ie. wie down) prety ‘what you want totestin an experiment [2] Devise a strategy. Record precisely how you can schieve the objective This includes thinking about the size and structure of the experiment ‘how many treatments? How many replicates? How will the results be analyzed) [B]Ser down all the operational details How will the experiment be performed in practic? In what onder wil things be done? should the ‘treatments be randomized or follow a set structure? Can the experiment be concluded within a time frame? Etc. ‘The point of any experiment isto determine the impact an explanatory variable has on a response variable. However, in any experiment, the following are usually involve: © The researcher ‘©The subjects (objects of the experiment) + Theevaluators 12 Statistical Scudy Page 2030stati sical studies include, * Observational st ee idies: In an ‘observational study, the researcher or * metely obseres what is happening or what has happened in “he pst and resto daw concson hase on these beratons lustration: Approximately 450000 vasectomies are performed each year ‘nthe United States. this surgcal procedure for ‘contraception, the tube ‘Carrying sperm from the testicles is cut and tied. Several studies have been ‘conducted to analyze the relationship between vasectomles and prostate ‘eancet. The results of one such study by [Link] et al. appeared in the paper “A Retrospective Cohort Study of Vasectomy and Prostate Cancer in US. Men (Jura ofthe Aner Medial Aston, Va 259) pp. 78-88, ‘Dr. Giovannucci, study leader and epidemiologist at Harvard affiliated Brigham and Women’s Hospital, sid that ~.we found 113 cases of prostate ‘cancer among 22.000 men who hada vasectomy. This compares to rat of 70 cases pet 22000 amongimen who didnt havea vasectomy.” The study shows about a 6 clvated tsk of prostate cance for men who have ada vasectomy, thereby eveang an asociation between vasectomy and prostate cancer, But does it establish causation: that having. a ‘vasectomy causes an increased rskof prostate cancer? ‘The answer is no, because the study was observational. The researchers simply observed two groups of men, one with vasetomies and he other without. Thus, although an association was established between vasectomy and prostate cance, te association might be due to other factors (eg, temperament) that make some men more likely to have ‘asectomies and also put them at greater risk of prostate cancer Page 30130,F a | + Experiment peti: ec al studies: In an experimental study, the researcher lates TUES one of the varales and ees to determine how the ‘aniptlation influences other variables, Uh ‘stration: For several years, evidence had been mounting that folie aid ‘educes major birth defects Ds. AE Czezel and Dudas of the National Tnstirte of Hygiene in Budapest decd study that provided the Strongest evidence to date, The els were published in the paper “Prevention of the First Occurrence of Neural-Tube Defects. by Periconceptional Vitamin Supplementation” (New England Journal of Medicine, Vol. 327(26),p 1832) For the sud, the doctors enaled 4753 women porto conception and divided them randomly into two groups. One group took daily soulkvtamins containing 08 mg of fle aid, whereas the other group received only trace elements (minute amounts of coppet, manganese, zine, and vitamin C). A drastic redeton inthe rate of major bith defects ‘occurred among the women who took folic acid: 13 per 1000, as compared ‘to 23 per 1000 for those women who did not take folic acid. In contrast to the observational study above this isa designed experiment ‘and does help establish causation. The researchers did not simply observe ‘wo groups of women bu, instead, randomly assigned one group to take daily doses of fli acid andthe other group to take only trace elements. Definition of some terminologies used in the Design of Experiments Experimental design and analysis has a standardized terminology that is, ‘unfortunately, different from that used in survey sampling, ‘+ Factor: one ae under the contol of the experimenter that is ‘varied over different experimental units If variable is kept constant over 13 all experimental nits, then iti not a factor because We cannot discer its influence on the response (¢g ifthe water temperature in an experiment i Page 40130,puke ‘equal for all tanks, it is impossible to determine a t), Factors are sometimes called explanatory variables. A factor has 2 oF more levels levels: values ofthe factor used inthe experiment, For example, in an ‘experiment to assess the eects of different amounts of Ulra Violet (UV) radiation upon the growth rate of smal, the UV radiation was held at Dorma, 12 normal, and 15 normal levels, These would be ee tree levels for this factor. Treatment: the combination of factor levels applied to an experimental unit. If an experiment has a single facto, chen each treatment would correspond to one ofthe levels. fan experiment had two or more factors, then the combination of levels from each factor applied to an experimental unit would be the reatment. Fr example, with two factors having 2 apd 3 levels respectively, there ae 6 possible treatment cominations Response variable: the outcome that is being measured. For example, in an experiment to measure smolt growth in response t UV levels, the response variable foreach smolt could be fina weight after 30 days. Experimental unit: the item to which the treatment is applied. For cxample, several smolt could be placed into a tank and he tank is exposed. to diferent amount of UV radiation. The tank is the experimental unit, In agricultural experiment, the experimental unit may be pots of land and also in linia trial the patients could be the experimental units. ‘Blocking: this is a process by which experimental unit is partitioned into groups of homogeneous units. It is intended to ensure reduction im experimental ertor by isolating all possible source of variation, Blocking is thus a technique used in increasing the precision ofan experiment. Page of 30EE EXPERIMENTAL DESIGN An experimer ntal design i Ttis used i is It is used in all cee the arrangement of treatments in an experiment, is more diffi empirical investigati Jignit i at es teas gations. Designing of an experiment (part of planning) and a “speci its analysis, because it means both designing of an experiment to worn ic design.” Finney (1955) defined the “design” (i) The set of t (a The a selected for comparisons its ‘on of the experimental unit (animals, ficld plots) to which 7 eareyer are to be applied a ; rules for allocating the treatments to the plots oF units iv) oe) specification of the measurements or other to be made on unit (v) Specification of details of the management of the experiment ‘The need for a statistical subject of design and analysis of eat? wre from variability inherent O16 experimental material environment ‘and management. This ‘variability leads tO mental ended = lable variability in the "ocrved variate values Hy 1° experimental eFF0F” qrhere “error” does nol mest “gnistake ‘The four major components of design (1982) are as follows: (a) Planning, design, and layout (b) ‘Management (@) Data record or data (a) Serutiny 2m ting. We describe pelow the implications of the four components listed above page 60130PLANNING, DESIGN AND LAYOUT ‘This component includes the following i) A statement " Viguote atthe objectives ofthe experiment. A ler and unan- pa hypothesis must be formulated. The investigator must be very ‘ATi with the subject matter, and thorough iterate search nd Teen reh eer a thn em of erro imrtnc s ing the knowledge needed to formulate a testable hypothesis. The ‘question you are trying to answer should be very specific and clearly ‘tated. Examples of hypotheses are listed below: (a) Alternative Hypothesis ~ Groups are expected to show different results, eg. rats will gain ‘more weight on diet A rather than diet B. (b) Null Hypothesis ~ Groups are expected to be the same, eg. rats on diet A will xin the same amount of weight as rats on dict B. (©) Untestable = A result can not be easily defined or determined, eg, rats on diet ‘A-will look better than rats on diet B. What does better mean’ AA definition of “better” must be clearly stated. ‘Once the question and hypothesis have been stated, the methods and ‘techniques to be used can be determined, and evaluated for the best possible method to perform the research. }) Definition of the population bout which the inferences are to be made. ) Selection of experimental treatments, which must include a control treatment/s. Sometimes « strictly “untreated” control may seem ap propriate, and sometines in medical vocabulary, a “placebo” Controls generally take four different forms viz., negative, comparative, (iv) Choosing the plot shape and size, ‘Choices of (ji) and (jv) must be guided by the principles of replication, randomization, and blocking (local control). ‘These concepts will be fully discussed in section 9.3. vehicle, positive, and age 7 130MANAGEMENT. Administration, supers atic care if reich ns 8M management ofan experiment ned eystem- ais results are to be obtained. If plots are pieces of land, they rates of deal UP ad planted propery. Ifthe treatments are different Can os these must be determined ctrl applied, Ene tke to avoid extra variability after treatments have been . Extra variability can be introduced into the results by different man- ‘Bement of the plots, eg, by having some plots weeded more carefully or frequently than others in afield experiment where treatment differences are not intended to include weeding differences. Such circumstances may require Afferent blocks to be weeded by different people, so that unwanted between- laborer differences are assimilated in between-black differences. Likewise, if an ‘experiment has to be irrigated throughout, but there is insufficient equipment ‘irrigate it all at once, the irigation should be “by blocks.”. Also included in this category are, operations such as harvesting or scoring for diseases. DATA RECORDING ‘There is a need for suitable balance for obtaining accurate, precise data values. When weights, heights, and other continuous variables are to be recorded, a “degree of precision of recording’ should be specified and ad- hhered to, eg, “to the nearest 5g." “to the nearest 25 em," ete. Data should be recorded on prepared datasheets having proper headings. Copying of data introduces errors and should be avoided if possible. SCRUTINY AND EDITING OF DATA ‘This involves searching for outliers. Quality assurance procedures to identify data entry errors should be developed and incorporated into the experimental design before data analysis 4 Principles of Experimental Design “There are chree (3) important principles inherent in all experimental designs that are essential tothe objective of statistical research; Page Bo 30Randomizati tion: : oe a Randomization is the assignment of treatments to r Nal units so that all units considered have equal chance of ‘Scelvng a treatment. It ensures unbiased estimates of teatment means and experimental err, “Replication: Replation means that a treatments epeaed two ore ‘mes Its function is to provide an estimate of experimental ertot and to Drovidea more precise measure of treatment effects i to reduce the impact ‘frandomization ertor. The numberof replications that wil be reqited in a pticular experiment depends upon the magnitude ofthe diferences you ‘wish to detect and the variability of data you are working with. Considering these two things at che beginning of an experiment wil sive ‘nue frustration ‘ik Local contro: this is a process which includes all practices or techniques ‘sed to minimize the experimental err such as balancing, blocking, and ‘srouping of experimental units. ‘Basic Statistical Design of Experiments 1. Completely Randomized Designs (One-way or Single factor designs) ‘A completely randomized design (CRD) is one where the treatments are assigned completely at random so that exch experimental unit ha the same ‘chance of receiving any on treatment. For the CRD, any diference among experimental units receiving the same creatment is considered as experimental ero. Hence, CRD is appropriate ony for experiments with homogeneous experimental units, such as laboratory experiments, where cavironmenta effects are relatively easy to control. For field experiments, where there is generally large variation among experimental plots in such environmental factor ssi, the CRD is rarely used Page of 30Data structure for CRD i ee ae 8, = Trane Observations | 2 ; 7 te Sherine ag a oe : ym Ya . wy m : Jy Teal Ys Yn ” wy Ye » ‘Table I: Data structure for equally replicated CRD “Treatment Observaons. 12 gn) Totals eer ts v mn mi mt 2 mn ee ' SP pata Ye Tol ya Ya 71 tare avis “Table 2: Data structure for unequally replicated CRD ‘he t eaments are aanged in columns () and cbservation levels due vo numberof treatment repiatons ar in ros (). The CRD can either be of equally replated (balanced) or unequally replicated (unbalanced) types, respectively. The ‘hic all ehe treatments ae replicated equal numberof times (1). Henee, fy = Ta = formerisone in 1 = and otal umber of experimental lots or observations (ithe product of the ‘number of treatment (1) and the numberof replications ( (jythaeis n= rt, Thelatteris ‘oe in which at lastone ofthe j treatments ha different number of replication fom others Hence, there ests # ry for j = 1.2.3 ~The total numberof experimental plats or observation (ns 3 Page 10030,‘Atvanags of completly randomized desig L Cony ae Axil is allowed ~ any numberof eatments and repliates may be 2 Relatively sy statistic analysis, even with variable replicates and varable xPerimental errors for different treatments, 3Analyss remains simple when data ae missing 4: Provides the maximum numberof degres of freedom for eror fora given number ‘of experimental units and treatments. Disadvantages of completely randomized designs | Relatively low accuracy due to lack of restrictions which allows environmental ‘variation to enter experimental ettor. 2 For large numbers of treatments, a relatively large amount of experimental ‘material is needed which increases the variation. Appropriate use of completely randomized designs 1. Under conditions where the experimental material is homogeneous, 16. laboratory or growth chamber experiments. There is no satisfactory basis for ‘arctioning the experimental material into homogeneous portions. Where a fraction of the experimental units i likely ro be destroyed or fail to respond In small experiments where there isa small numberof degrees of freedom. Ieissutable for any nomber of eatments and any number f experimental units Ichas only one systematic source of variation such that there is high error degree of freedom. 6, Teeanbe equally or unequally replicated, ANALYSIS OF VARIANCE (ANOVA) FOR CRD” [Link] method of analyzing experimental data is known as ANOVA technique. The procedure attempts to analyze the total variation of a response by decomposing it into independent and meaningful portions attributable to each ofthe independent variables Page i of 30and to chance vari ee a ‘The ANOVA mode arising from the typical data structure for an Picsted CRD asllustaedin table above given by Weta rey Observation é arising from the effect of treatment j H= Overilllorgrand mean = Effect of treatment j ey = _® Errorcomponent assumed to be normaly, identically and independently disebuted with mean zero and constant but unknown variance ie. ey~N(,0"). ANOVA TABLE [Sourceof [Degree of] Sumof | Mean square] Computed Tabulated | Variation | freedom (DF) | squares | (MS) F (Fratio) \ey) (Ss) | treme | e-1 Sst pa \ MSE ee | i Total | ani ‘SST Ween rt (CFisthe corection factor, obtain as, (Gorequal replication) Page 120130,(Gorunequal repiaion) Coefficient of Variation (CV) ‘The CV affects the degrce of precision with which the reatments are compared and is good index of the reliability of the experiment. It is an expression ‘of the overall experimental error as percentage of the overall mean; thus, the higher the CV value, the lower is the reliability of the experiment. The CV varies greatly with the type of ‘experiment, the crop grown, and the characters measured. An experienced researcher, Ihowever, can make a reasonably good judgment on the acceptability of a particular CV ‘value for a given type of experiment. Experimental results having a CV value of more than 30% ate toe viewed with ution SE Grand Mean Where, x 100 Grand wean 2 ANOVA Test of significance In this section, procedures for hypothesis test of significance of difference between treatment means called ANOVA F-test willbe sed in testing significance, ANOVA test {s said to e significant if and only the null hypothesis of equality of treatment means (Ot treatment elects) i rejected based on the test criteria, otherwise its not significant “The hypotheses tobe tested in ANOVA settings are stated as follows, Null hypothesis (Ha): all treatment means are the sume or cffects of all treatments are the same 2 age 13 0f 30 Slater A"| Meative hypothesis ‘Mesmate hype (i atleast one treatment means diferent fom anther oF eflect of atleast one treatment fers from another Symbol Hy y= HyNot Hy Thetest statistics giveninthe ANOVA ale hove as i Plas ee sa ae = SE Decision Rule: Reject My in favour of Hy if Frais Fans athervise, do mot reject Ho oF Reject He in favour of Hi P-value < a, otherwise do not reject Hy Decision: Decide-based on the decision rule CConelusion: conclude based on your decision Example ‘The table below shows the outcome ofa laboratory experiment performed by a professor of microbiology at Kwara state university, male, in which observations were made on ynycelal growth of diferent Riytona sola isolates on Potato Dextose Agar (PDA) 2A um The data represents Myc groin terms of Aameter ofthe sly (mm) of R solani isolates on PDA medium after M4 hours of incubation Ticks BS BS? SS RSH_BSS oe T Bs 5 SMS xe awa 2 BMS M465 kya’ 3 » 9 49 set =< Gi Toul as OSES OSS Att 4ee ain the flowing 4) Factorinvolveinthe experiment x. @*Sans’ by Number offactor eels ©) Experimental unit wh Number of experimental plot ©) Number of treatments 4) Type oreplication involve 8) Response or dependent variable =, thyalin gout bh) ANOVA table 1) Significance of treatment effects t 5% eve of significance 4) Coefficient of Variation SOLUTION wf _ G46 a2 AO 30123 cr a 183 2 sv=)4-cr (a6)? 4? | (565)? | (114)? 655"] fo 00 8500 881 = 76269 sty oF fist = (C29)! + (28)! + (29)? +--+ (B1)4) - 15301.23, =78927 Pages of 30(hy ANOVA table for example SumofY Meamsquare [Computed] Tabulated a Ms F(Fraua) | (Fut) sett re) Diselan 72. Doe Te] ae O:DRE 2658 332 789.27 ie (0) Significance test Hsellects of all treatments (Rizoctonia solani isolates levels) on Potato Dextrose Agar(PDA) medium ae the same Higelfects of atleast one treatment (Rieti solan isolates level) on Potato Dextrose Agar(PDA) medium difrs Decision Rule Reject Hy in our of Hy Frag 2 Fay otherse donot reject Hy Decision Since Fais(5738) 2 Fp (3.04), we therefore reject Hin favour of Hy ‘Conclusion: we therefore conclude that eto a east one treatment (Rgxtonasolan isolates eve) on Potato DestroseAga(PDA) medium differs from another. (). Coefficient of variation (CV) ‘WE lh nee Where, ‘Thisindicates thatthe experiment is very reliable ata value of 94.69% Page 6030 Praue Sag up A Pale as hig Yor tee vale Aon be Nite NetoANOVA Model Parameters Thetwo pan FMtumetes ofthe ANOVA model above under a CRD are wand respectively The Least ure estimator derail fomalas fortes parameters ate expressed 38 1 een efi SLAVE 5% 9 (forequalrelicatiod Bale pHi 25. (forunequal repl. icatiod and Ea f=3)-3, (forequalreplicatiod 5, (Forunequal replication 565 446 ays = SF = 825 - 3431 MAO 6 ag = =p = 480-3431 = 1369 6 46 anit = SAAR 3072-3431 = 159 ‘Variances nd standard err of treatments means Frany treatment with mean J, th variance and standard error (Sf the treatment mean ate computed as 1) = waFe EE Example Using the AN ths ANOVA table obtain in example | above, compute the standard ero forall the treatment means ; s.e@,)= me P= Bsa FE - FB -1os For any two treatments j and j with mean Fj and J respectively, such that # j'-The variance and standard error (Sc) of the different between the treatment means (5) 3) are computed as 16,3) =MSELE +4 Foronepal repiaton ae sebys-9)* see] 2MsE VIF) == Forel MSE replication S.e(9j- Fy) = Exercise age 38030the ANOVA table obtained rumen pears ‘mexample Ito compute the standard error forthe differen between all = iets resend Conieton) ven the ANOVA F:te not the }OVA E-test is significant (His rejected), i indicates chat the treatments same, on the average, This does not necessarily imply that all the treatment ans are significantly different To determine which pair(s) i significantly diferent iewise multiple have to co Ne to conduct alte! comparisons called post-hoc test oF pal parison test. We would therefore need to make () comparison se ofthe most commonly used test procedures for pair comparisons ini the Fishers st significant difference (LSD) test. Other test procedures, such as Duncan's multiple age test (DMRT), the honestly significant dillerence (HSD) test and che Student swman Keuls range tstete can also be used LSD test is the simplest of the procedures for making pair comparisons, The ‘cedure provides fora single LSD value, a a prescribed level of significance, which ves as the boundary between significant and non-significant difference between any cans, Tat is, two treatments are declared significantly different 2¢ 8 ir of treatment mé eds the computed LSD scribed level of significane if their (Absolute) diference exces tue otherwise they ae nc considered sigofcantl ferent. The LSD procelure #8 lows: any two treatments and with mean J, and J, respectively, suc chat * ji The pothesis to be tested is say = hy (Treatments j andj are not sigifcanly ilferent) uj # ty (Treatments jad’ are significantly diferent) se statistic: D= 3.4 fyy % 2 ~FV) MSE For unequal replication ty, fer Page 19 of 30i * Stare ® PE For equal replication Decision Rule: Reject Hin favour of Hyat aleve of significance if |p Fy 2 USD. otherwise. do not reject Example From cxample 1 above, we need to conduct multiple comparisons to determine the ‘reatments pairs that are significantly different at 5% level of significance. Since t= 5, we would make () comparisons. Q-sma Hey = py (RSL and RS? are not significantly different) Huy # lp (RS Land RS? are significantly different) Lsp=ts,x tooasa X 149 306 149 2344 Ws ~ Fal = [2867 — 31.331 = 266 Decision: since 17s —F2\(2.66) < LSDB.A4) we do not reject Ho, We therefore conclude that (sreatments) RSL and RS2 are not significantly diferent, “The procedure is then prepared fo the other 9 paired comparisons. Class work (Obtain the result fr other 9 paired comparisons. EXERCISE page 20 0190 4 eee1. Four chemists (ay Abotatory scientists) were asked to detem ethyl alcohol in a certain chemical, maccae compound and the results ae as follows: Chemise Percentage of Methyl Alcohol ! 499 3408 438 2 85.15 35.13 8438 3 an 8448 85.16 4 84.20 84.10 8455 ®) Write a suitable ANOVA mode forthe experiment above. 5) Idenify the treatments, types of replicate, number of repliate and the response ‘variable in the experiment. ©) Ate there differences among the treatments at 5% level of significance? 4) Obtain the standard error forall posible treatment means ©) Obtain the standard error for differences between the treatment means, 2. The vice chancellor of KWASU Prof. A. Natallah performs an experiment in which ‘the amount (6) of radon release in showers was investigated. Radon ~ enriched ‘water was used in the experiment and different orifice diameters were tested in the shower heads. The data from the experiment are shown in the table below. Oise Pees eee 037 as) on 1 140 19 seoras Obtain the following 4) Factor fnvelvein the experiment > OTE daasestsr b) Number of factor levels — ©) Experimental unit — Roaches gmmeb~ a) Number of experimental plot — 244 ¢) Number of treatments _ ok pale Page 21 of 30 i) Typeof replication invalye — 1) Response or dependent variable h) ANOVA table i) Si gnfiance of treatment effects at 5 evel of significance 1) Coeicent of Variation forthe experiment 4) Obtain the standard eror forall possible treatment means ‘) Obtain the standard erro for differences betwen the treatment means RANDOMIZED COMPLETE BLOCK DESIGN ‘The randomized complete block design (RCBD) is one of the most widely used ‘experimental designs in research. The design is especially suited for fel experiments where the number of treatments isnot age and there exit a conspicuous factor ase ‘on which homogenous sets of experimental units can be identi xd. The primary isinguishing eat of the RCBD is the presence of blocs of equal size cach of which contains all the treatments, The purpose of blocking sto reduce the experimental e=FOF by eliminating the contribution of known sources of variation among nits Thiss done by grouping che experimental unis into blocks such that vary within each block is minimized and variability among blocs is maximized. S the variation within a block becomes part ofthe experimental error, blocking is most effective when the experimental area hasa predictable patter of variability the experimental ince only Whenever blocking is used the identity ofthe blocks and the purpose for their ‘be consistent throughout the experiment. That is, whengve that is beyond the conttal of ‘ase must sa source of variation exists the researcher, it should be ensured that such variation ‘occurs among blocks rather than within blocks. For exam ple if certain operations such as appliation of insecticides or data callestion cannot be completed for the whote experiment in one day, the task should be completed for all plots ofthe same block: 2” the same day. In this way, variation among days (which may be enhanced ‘by weather factors) becomes a part of block variation and is, thus, exelud led from the experimental Page 220f 30 Percrtaye gerout of: vetdors ‘altacebe staal Soh make measurements oral pos ofthe came variation among observers if any, would con block. This way, the deena i stitute apart of block variation instead of Basic Features of RCBD 1 lkisnotas flexible asthe CRD ‘is suitable when the experimental material is heterogencous and the suspected heterogeneity operates in one direction. 3 Incomplete randomization of treatments to experimental units is adopted in blocks. ‘The number of treatments t is equal to the size of each block ie. t k. [As aconsequence ofthis all treatments are replicated the same number oftimes It has two systematic sources of variation due to treatments and blocks respectively. Data structure for RCBD, ben hey 2 eh ee Teammate es en Blocks a + To ea z se Tm a va ye 2 ete Ya me t Number of treatments ‘umber of blocks Total (n) = bt Page 230130NOVA Model for RCRD ‘he appropriate ANOY, ‘Atmodelin RCBDis given as, YW FHt B+) +e, Observation in block arising rom the effect of treatment J H = Overalorgrand mean B= Effect of block é 4 = Efleccof treatment y= _ i Errorcomponent assumed to be normally, identically and Independently distributed with mean zero and constant but ‘unknown variance te, ey~N(0,02). ANOVA Model Parameters The parameters of the ANOVA model above under a RCBD ae 4, and ay. The Least square estimator dervational formulas for these parameters are expressed as Ver ZierFu Ye mre ANOVA table for RCBD Page 240f30i a0 |p, vai (DF) | squares (65) 0) ‘Compated F | Tabulated F rs Fraie) — | (Fas) fireatment t-1 Sst ak leat EY Far-no-ne-n os ea 5B co Error ee {feats O-De-9] ssp [uses t ry Toul beat cr med | ee de el ee Where ssraSSyh- co A jat SSE = SST ~SSB-SSt Example 2 “Thedata below isthe outcome of an experiment conducted by a student in the Faculty of Agricultural Sciences atthe better by far university (UNILORIN). The experiment was designed to investigate the ferences in yield of wheat (tonslaces) for seven varieties of Fertilizers. The experimental field was divided in to five blocks, each containing sever. cxperimental plots. The seven plots were assigned fandomly to the different fertilizers, ron we’ one plot for each fertilizer anda yield was recorded foreach plot Page25 of 30