Introduction:The phrase’longitudinal’ is used to symbolize a set of studies that are conducted over aperiod of time. Often, as we have seen, the word ‘developmental’ is employed inconnection with longitudinal studies that deal specifically with aspects ofhuman growth (30)Longitudinalstudies (defined in general as studies in which the response of each individualis noticed on two or more occasions) represent one of the principle researchstrategies employed in medical and social science research (Goldstein 1979;Nesselroade and Baltes 1979). Longitudinal designs are uniquely convenient tothe study of individual change over time, involving the effects of development,aging, and other factors that affect change. Despite the importance of thelongitudinal study, however, satisfactory methods for the analysis of serial measurementsare not readily available. The statistical literature on the analysis of serialmeasurements is based on the paradigm of multivariate regression, and standardstatistical software packages have the same orientation. Yet longitudinalstudies typically have Unbalanced designs, missing data, attrition,time-varying co- varieties, and other characteristics that make standardmultivariate procedures inapplicable.
(21)Longitudinalstudies have a long history in medical and social scienceResearch. Theyoffer a natural access to the study of development andAging thatallows the separation of age and group effects. They can also beUsed toproduce precise estimates of treatment contrasts not subject toBetween-individualvariability. Yet they are often more difficult and necessitateGreater spendingper surveillance than cross-sectional studies. (21)The momentthat we have decided to running a longitudinal investigation, we mustMark exactlyhow, when, where, and on whom the measurements willBe taken.
Longitudinal designs principally have subjects crossed withOccasions.This means that we must choose designs for picking out subjectsAnd also theoccasions at which they will be measured. (21)The designof a longitudinal study has two aspects: a design for selectingSubjects anda design for occasions. Typicallongitudinal research is characterized by:The obviousintegration of three elements: (a) a well-expelled theoretical model ofChangeobserved using (b) a temporal design that bears a clear and itemizedview ofThe process,with the resulting data analyzed by means of (c) a statistical samplethatIs anoperationalization of the theoretical model? Two general varieties oftheoreticalModels areconsidered: models in which the time-related change of primary interest isContinuousand those in which it is characterized by movement between unattached states.(24)One of thescientific benefits of directing longitudinal study is the ability to observetemporal order of a key exposure and outcome events ,particularly we can locatewhat ever changes in a covariate precedes changes in the outcome of theinterest (25). A possible hardnessin longitudinal study is that the measurement of defendant may be missing atone or more time points (32)The longer astudy keeps the higher will be the value of the resulting longitudinallarge-scaleData sets.The clear reason is the permanent collection of information concerningLife-coursedevelopments.
(8)Longitudinaldesigns have two primary motivations:1. To raisethe regulation of treatment contrasts by eliminating interindividualVariation:This is obtained by watching each subject under theSeveraltreatment (or exposure) situations to be compared. Such designsAre calledrepeated measures designs, and involve the cross-over designAs a specialcase.
Repeated measures designs use each subject as his or,Her owncontrol.2. To testthe individual’s coverting response over time: LongitudinalDesigns havenatural entreaty for the study of changes correlated withDevelopmentor aging. They have value for describing both temporalChanges andtheir depending on individual characteristics.(22)new researchhas focused on the development of statistical methods that not only take intoaccount the inter- correlation of serial measurements but also harmonize thecomplexities of typical longitudinal data sets and declare the specification ofmean-value functions determined by subject matter considerations rather than byconstraints introduced by the statistical methodology.
This work has been basedon the concept that the analysis of serial measurements should be inspected asa univariate regression analysis of responses with correlated errors. (21)Longitudinalstudies are distinguished by frequent observation of individual respondents.(The terms respondent and study participant are used because of our primary assuranceon human investigation. The methods described, however, apply quite generallyto longitudinal research, and the term experimental unit has the same technicalmeaning.) The metameter for the occasions of measurement may be age, time onstudy, or some other natural scale; alternatively, the relations of measurementmay match to levels of an experimental or observational variable, possibly withno natural order in. (21)The excellencein longitudinal research requires care in both theDesign and attitudeof studies. The general goal in conducting the study isTo performthe study as designed.
Three important objectives in performingLongitudinalstudies are the following:1. Bringdown and quantify instrument and observer variability.2. Guaranteerigorous recording, coding, and transcription of data.3. Bringdown nonparticipation and obtain complete data at regular visits (23)Longitudinalstudies give more precise rating of temporal changesOr treatmenteffects than cross-sectional studies of the same size.They attainthis gain in precision by completely removing interindividual variability fromthe comparisons of interest. DISCUSSION: A studydesign is a project for a data set that will efficiently test studyHypothesesand evaluate important parameters.
Since analyses depend onThe dataactually collected; however, the importance of a study depends on equallyOn itsdesign and its implementation. Although the details of a well-conductedStudy basedon the variables measured and the goals of the study, someRequirementsfor a well-executed study cut across disciplinary lines.In any studyinvolving measurement, sources of bias and measurementError mustbe controlled. Generally, this requires inclusive training ofObserversand evaluation of the validation and reliability of measurementTools, forexample, in the Hypertension Detection and Follow-up ProgramA study inwhich blood pressure level was a critical measure of treatmentEffect,random zero sphygmomanometers were used to avoid bias.Arising fromdigit preference or other subjective factors, Blood pressureMeasurementdevices were rigorously pretested, and observers were neededToparticipate in a one week training course and pass a certificationExamination,these procedures were intended to minimize interobserverAndinterinstrument variability.In alongitudinal study, there is an additional need to maintain studyProceduresover time, Instrument performance must be constantly watchedAndsystematic bias re-estimated.
Similarly, investigators should planFor trainingof new observers and recertification of settled observers toBypassdeterioration of measurement procedures. Because this element of aStudy doesnot participate directly to study results, it is a natural candidateFor reducedeffort during periods of tight budgets. To avoid this mistake,Studydesigns and study budgets should include sufficient support for thisWork, Investigators should set a regular schedule for checkinginstrumentsAnd observersand the results of these checks should be kept asA part ofthe study records. When proper, instrument and observerNumbershould be recorded with each observation, so that these variablesCan be regardedin the analysis if necessary, In some studies, instrumentAnd observervariability are comparable in significance to the effects underStudy andmay bias estimates of effects of interest if not cautiously controlled.Bias canalso arise in the recording, transmission, and entry of study data.Althoughmany studies confirms extensive quality control for data entryTo computerfiles, variability in reading participant records such as Spiro gramsOr coronaryangiograms and inconsistency in coding questionnaireData oftenrepresent more important sources of variability.
These sourcesOfuncertainty can be quantified and controlled only by extra qualityAssuranceactivity in the data collection system, although observer variabilityIs a widely knownphenomenon, specialists are often surprised by theAmplitude ofobserver variability in interpreting diagnostic tests such aselectrocardiograms,Thus,duplicate readings should be introduced to quantifyThis variabilityand blinded reading should be required when there is potentialFor bias incomparative work,Awell-designed longitudinal study can also be intimidated by problemsOf missingdata, Data can be missing either because of procedural errorDuring aregular visit or because a participant does not appear for a regularVisit. Bothevents are harmful to the study. Most procedures for treatingMissing datain the analysis suppose that data are missing at random (44),I.e.
theprobability of missing an observation does not depend on the valueof thatobservation. When that hypothesis is true, the main outcomes of missing dataare (a) obstacle, because unbalanced data sets areMoredifficult to analyze, (b) loss of regulation, because missing outcomesDecrease theeffective size of the study, and (c) problems with modification forCovariateswhen their values are missing.Inlongitudinal studies, the proposition that observations are missing atRandom isfrequently not reasonable. Participants who are lost to follow-up areOftenatypical in terms of mobility, social class, and general health. This isA specialthreat to comparative studies in which different groups haveDifferentfollow-up processes, this problem is overcomes by introducingExtraprocedures to control follow-up of study participants and promoteParticipationat regular visits.(22)Prospectivelongitudinal studies are studies which include repeated measurementOf the sameindividuals over time. Panel analyses are an example.
Experiments, inWhich aparticular „treatment** is given to a portion of the study groupBetweenrepeated measures of the same subjects, could in principle be seen as ofLongitudinaldesign as well. However, as long as the before and after design involvesA short timeperiod only, we excel not to include experiments. Prospective longitu¬dinalstudies might take on various forms according to their variables: If the sameIndividualsare repeatedly measured using the same variables the term panel analysisIs usuallyapplied. If the same individual is repeatedly measured using differentVariables,the term prediction studies are not uncommon. Both kinds of longitudinalAnalysismix, since even in prediction studies, conceding marital adjustmentOrdelinquency for instance, some variables remain the same in the subsequentMeasurementwaves. possibly for this reason, many authors see the terms panel andLongitudinalstudy as equivalent Prospectivelongitudinal studies might also take on different forms with regardTo the timedimension of sampling.
One may start from data collected in the pastBy otherresearchers and follow them up. One could, of course, also start from thePresent and forwardto the past for given individuals by taking recourse to archivalData. Inretrospective or quasi-longitudinal studies the subjects have only one measure¬ment in timebut data attaching to various points in time , decisions and definitions theybrought to the Situation in past times.Thelongitudinal approach in this design is thus one depended on the memoryOf therespondents, various data collection strategies have been associated with thisApproach,ranging from more or less unstructured forms of data collection (such asinqualitative analysis of specially triggered written autobiographies)To highlystructured closed interviewing Samples have accordingly varied be¬tween samesamples of special social groupings to large samples of the population inGeneral. Wherelongitudinal studies are started or continued the most significant practicalProblem mostlyto be solved is retrieving the respondents. Depending on the char¬acteristicsof the population, the records with which one is starting, and the timethat has passed,the number of cases which manifest to be lost at the outset is nor¬mally very high.
There can beno doubt, nevertheless, that some studies — according to sample and de¬signcharacteristics — have bigger chances than others to succeed, regardless of thetechniquesused to determine respondents. Leading factors affecting tracing failures arethe size,mobility and scuttle of the sample. Among social groups relatively ho¬mogeneousand centrally located populations with high education (such as universi¬ty students)the researcher probably fairs best.
The chance of locating them mightbe highsimply because people tend to stay in these institutions for some time,establishcontacts with many people there, and later follow certain relatively homo¬geneous careerpatterns. They might, moreover, belong to alumni Organization andToprofessional organizations, whose directories might be searched. High educationMightfurthermore raises cooperativeness.(40) (34)To analyzelongitudinal data, one must appoint the probability distributionfor eachsubject’s set of responses. We shall at first suppose that the outcomeVariable isa measurement that has a normal distribution, The approachto modeling the predictable outcome depends on the goalsof thestudy. In repeated measures studies, differences between occasions inthe expectedoutcome for a single subject are refer to changes inTreatment orexposure conditions, the analysis starts by testing the equalityof meanvalues over occasions and continues with estimation of theDifferencesin means between occasions. When changesin the predictable outcome over occasions are due to growthor aging.
orwhen occasions coincide to different levels of exposure, theanalyst willwant to model the changes over occasions. (22)Unplannedmissing data in longitudinal research can occur because a participantFails torespond to one or more questions in a questionnaire or interview, or becauseaparticipant is not obtained to the research study at one or more opportunity ofmeasurement.Scientists involved in longitudinal research deal with unplannedmissing datapermanently; in fact, it is difficult to visualize a longitudinal study withoutat leastsome unplanned missing data. For this reason, how to handle missing datais animportant question meeting anyone who wants to analyze longitudinal data.For yearsinvestigators have used ad hoc procedures for dealing with missingdata, suchas eliminating individuals with missing data from analysis (“casewisedeletion”)or substituting the sample mean for missing observations (“mean substitution)Suchprocedures may be adequate, but they have no basis in statistical theory.
Thereare two prospect consequences of using ad hoc procedures to dealwithunplanned missing-ness. One is a higher-than-necessary loss of statisticalpower,particularly in association with case-wise deletion, which including discardingdata for anysubject whose data are incomplete. The other consequence is biasin results,which can occur if the cause of missing-ness is linked to variables ofscientificinterest.A much advancedoption is to use modern missing data procedures (Schafer 1997,Schafer& Graham 2002), such as multiple imputation and maximum likelihood,which are dependson statistical theory.
When the assumptions underlying these proceduresare met,they restore much statistical power and eliminate bias due tomissingdata; even when the underlying assumptions are not met, modern missingdataprocedures are an improvement over ad hoc methods (Collins et al. 2001).This isparticularly so if variables that are highly correlated with those subjectto missing-nessare involved in the analysis. Commonly in longitudinal studies,Past orlater measures of a variable may render this role well. Collins et al.
(2001) andGraham (2003) clarified why and how to implement this the better approach, and pretendhow such a sample can be tremendouslyprofitablefor making the most of modern missing data procedures in longitudinalresearch.(24)The longer astudy lasts the higher will be the value of the resulting longitudinal large-scaledata sets.The clear reason is the permanent accumulation of information concerninglife-coursedevelopments. Moreover, in educational research there is a obvious needforlongitudinal large-scale assessments because competence development andeducationalpaths can bewatched out for longer time spans.
Long-lasting studies are surelychallenging,even if a lot of time and resources are invested in keeping a panel stable andindiminishing biasing effects. Therefore, the design of longitudinal large-scalepanelsshould notremain fixed over the whole course of the study but should be adapted fromtime totime. In this direction two main levers exist. The time spans betweeninterviewscan bewidened basically, possibly with panel care measures in between. Alternatively,a panel canbe discontinued in order to invest the resources in a restart of a new panelwith a freshand unbiased sample.
This is especially valuable when, for example, fundamentalchanges haveoccurred in society or in settings that are of study interest, andwhich wouldotherwise be insufficiently covered.(8)in spite ofthe fact that missing data has the chance to cause serious bias, it is stillpossible to carry out a adequate and rational analysis. In order to do so,however, interest must be given at each stage of the research: design, datacollection, statistical analysis and reporting. in order to address the issueof missing data in patient recording outcome, including key definitions,prevention practices, and analytical approaches, including sensitivityanalyses.
We will not discuss questionnaire development, and make the tacit suppositionthat the patient recording outcome used is psychometrically dependable andvalid in the target population.(44)While somemissing data in longitudinal studies is nearly unavoidable, there are ways tominimize it. The first step is literally, the first step: in the design.
Whether or not the patient record outcomes are a primary or secondary outcome,they need to be completely integrated into the design and the manner of thestudy, with this integration codified in the study protocol, quality assurancemeasures, and statistical analysis plan. One design decision that impacts onmissing data is the decision to continue assessments after the patient missesan assessment or goes off treatment. In some settings the effectiveness oftreatment on the outcome will occur after treatment failure. Continuedassessment is conservative; if it is decided later that these data are not closelyconnected to the research question they can be excluded. This testament isbalanced with a warning about the length of follow-up in populations with highrates of morbidity.
Assessments should generally not be planned after themedian survival (and possibly should be a shorter interval).(44)
