Indosearch Research Center [IRC]

It is very important to save data file when you are working with SPSS. SPSS will not save it for you. Then, you must do it manually and frequently enough in case thing goes wrong, such as a sudden blackout.

How to do it? Go to the File menu (top left hand corner) and choose Save. You can also just click on the icon that looks like a floppy disk which appears on the toolbar at the top, left of your screen. Or you can use shortcut by pressing Ctrl + s on your keyboard.

If you are working with a new data and has not been saved, SPSS will ask you to specify a name for the file and to indicate a directory and a folder that it will be stored in. Choose the directory and then type in a file name. SPSS will automatically provide all data file names the extension .sav

J. A. Keats
After a brief statement about measurement theory, this entry describes “weak true score theory,” its implications and difficulties in practice. This section is followed by short statements on guessing in multiple choice tests, item homogeneity and “ordinal true score theory.” The problems of weak true score theory gave rise to the development of “strong true score theory”, which is described in detail. Finally there is a short section on the future of true score theory.
Although measurements of different kinds, such as educational and psychological, have been used for many centuries in different cultures, it is only in this century that theories of measurement have been developed. This fact raises the question of the role of theory when so much practical application has existed without it. One answer could be that although theory had not been enunciated it was implicit in practice: for example, the need for accurate standards of measurement became evident when subjective methods of marking essay examinations were found to be unreliable.
In mathematics, the need for an axiomatic approach became widely recognized in the late nineteenth century, and was also prevalent in the sciences in relation to physical measurement. Campbell's influential book on the theory of measurement was a result of this movement (see Campbell, 1957). The practices of measurement could now be evaluated in terms of this theory. Full Reading ...

The challenge for educational research is to integrate the findings of the large body of research being undertaken n Where estimates are made of effects meta-analysis procedures can be used to combine effects obtained under different conditions, in different contexts, and in different studies. Full Reading ... PPT Download ...

Three Areas in Educational Research Pose Major Problems of Data Analysis:

• Learning and the measurement of change
• Organizations and the analysis of multilevel data involving schools, classrooms and students
• Meta analysis that combines the results from multiple findings from multiple studies. Full Reading ...

Lawson and Hogben (1994) examined the effect of the keyword method on learning the meanings of words. Pairs of students from two year 9 classes in the same school were matched on the basis of their scores on an ACER word knowledge test and randomly assigned to two groups.

The treatment

Both groups were required to learn the meanings of the same series of words. However, the control group used personal learning strategies while the experimental group was trained in the use of an elaborative keyword strategy. PPT Download ...

Maximum Likelihood Estimation

The maximum likelihood approach to estimation and statistical analyses was initially prepared by R.A. Fisher but remained largely unused and underdeveloped until computers become widely available. Lawley (1940) was tackled the problem in the field of factor analysis where the task was to find a factor matrix in the sample data most like that which existed in the population. He was the first to derive the necessary equations, but unfortunately the equations could not be solved by direct computation. Subsequently Lawley (1943) advanced an iterative solution for these likelihood equations, and Jöreskog (1967) showed how a computer could be employed to provide the iterative solution to the complex set of equations, based on the assumption that there is a particular number of factors. At the end of this analytical procedure a test is applied to examine whether the solution involving this number of factor is adequate to model the sample data.
Today the likelihood approach is the basis for modern statistics largely replacing the least square approach. The maximum likelihood procedures require knowledge of three components:

(a) a set of characteristics with relationships between these characteristics or even a single characteristics to be investigated;
(b) an underlying frequency distribution involved in the generation of the characteristics under investigation; and
(c) a set of observed data relating to the characteristics under investigation. The set of characteristics give rise to variables. There must also be a proposed mechanism that is considered to be a possible framework that accounts for the generation of the observed data for the variables under investigation. This proposed mechanism and its framework is referred to as a model that involves the variables, hypotheses and underlying frequency distribution, and which relates to the generation of the data.

The questions to be addressed by the data analyst using the maximum likelihood approach is whether the proposed mechanism and the associated model could have given rise to the observed data. A model can not be considered to be ‘true’ because an alternative model might always be advanced that better accounts for the observed data. However, a model can be accepted as adequate until proved otherwise, and can be employed in predication and as a basis for explanation. (J.P. Keeves) Full Reading ... Full Download ...

11. Specifying Models for Two-Level Data

Phenomena studied in social and behavioral research often have a hierarchical structure, where individuals define one level of observation and groups or social organisations define one or more higher levels of observation. In educational research, for example, there is an interest in determining effects of characteristics of the school, the teacher, and the teaching on the development of individual students. However, classrooms are nested within schools, and students are nested within classrooms, so the observational structure is unavoidably hierarchical. Hierarchical data structures are exceedingly difficult to analyze properly (Bock, 1989), and as yet there does not exist a fully developed methodology for how to analyze such data with structural equation modeling techniques (Hox, 1994). However, Muthén (1989, 1990, 1991, 1994) has shown how approximate maximum likelihood estimates of parameters in a two-level model may be obtained with standard software for structural equation modeling, such as Amos, LISREL and LISCOMP. The resulting model specification is
quite complex, however, so there have only been few applications of this approach so far (see, however, Gustafsson, 1997, 1998; Härnqvist, Gustafsson, Muthén, & Nelson, 1994; Muthén, 1990, 1991, 1994).
The MB language is, however, easily extended to allow two-level modeling, so with STREAMS two-level models are only marginally more difficult to specify and estimate than are ordinary one-level models. The recently presented Mplus program (Muthén & Muthén, 1998) also supports two-level structural equation modeling in an implementation of the same estimation principles as those used in STREAMS. However, STREAMS also supports the Mplus two-level model specification, so it is possible to take advantage of the general advantages of STREAMS (e. g., starting values and a common modeling environment) here too. The present chapter provides a self-contained description of the steps and procedures involved in preparing data for analysis, and in specifying and estimating two-level models. Full Reading ...

The analysis of variance, introduced by Sir Ronald Fisher near the beginning of the twentieth century, is widely used by behavioral and social scientists. As a class of statistical models, “ANOVA” provides a means for analyzing data that is both rigorous logically and mathematically, and sufficiently broad to address questions posed in a wide spectrum of investigations. This entry describes the range of different analysis of variance models, the questions they address, the types of data for which they are appropriate, and the logic by which they operate. Several newer developments and recent thinking about ANOVA procedures are described and demonstrated in an investigation of students' motivation.
The main function performed by ANOVA is to compare systematically the mean response levels of two or more independent groups of observations, or of a set of observations measured at two or more points in time. Analysis of variance techniques are validly applied in each of three types of investigations (Bock, 1975): (a) experiments, in which subjects are assigned at random to treatments determined by the investigator: (b) comparative studies, whose purpose is to describe differences among naturally occurring populations; and (c) surveys, in which the responses of subgroups of a single population are to be compared.
Consider several examples. First, an experiment might be conducted in which one randomly assigned group of students receives no formal spelling instruction, a second group receives 15 minutes of spelling instruction per day, and a third group 30 minutes per day. At the end of the experiment, the effectiveness of spelling instruction is assessed by comparing the mean scores of the three groups on a common spelling test. Second, a recent large-scale comparative study examined differences among the achievement levels of students attending American public schools, private schools with religious affiliation, and other private schools. Although many different analyses were employed, ANOVA provides the most direct comparison of mean achievement across the three types of institutions. Third, the International Association for the Evaluation of Educational Achievement (IEA) international surveys sampled children in such a way that the final sample represented the entire age cohort attending school within each participating country. Subjects within a country may be subclassified by their responses even after the data are collected, to examine differences in mean response le vel among subgroups of interest. In an example described below, Swedish 13-year olds are classified according to their fathers' occupations, and average levels of achievement motivation are compared using ANOVA. In a survey, unlike other types of investigations, removing the researcher-defined subclassifications still yields an intact population (all Swedish 13-year olds). Nevertheless, whenever mean comparisons are of concern, ANOVA remains an appropriate and powerful analytic tool.
The examples above are described as having one measured response variable. Studies that yield two or more interrelated response measures (i.e., multiple dependent variables) require multivariate analysis of variance (“MANOVA”) tests and estimates. For example, the spelling experiment might have included a subtest covering words that were explicitly part of the curriculum and another covering words of similar difficulty that were not taught. Achievement in three types of American schools may have been measured in terms of multiple subject areas (mathematics, reading, social studies, etc.) or in terms of both cognitive and affective outcomes.
In each case, summing the measures will obscure important differences among the subscales, while separate analyses of each scale may give contradictory or confusing results and limit the replicability of the investigation. Appropriate MANOVA procedures maintain the integrity of the original measures while providing tests and estimates that pertain to the set of responses jointly. Repeated measures analysis is employed in investigations in which the same subjects are observed at two or more points in time or under two or more experimental conditions. These include pretest-post-test studies or educational or social interventions, longitudinal studies in which data are collected on the same scale repeatedly, and within-subject experiments in which subjects are measured on the same response variable under several experimental conditions. Both univariate and multivariate analysis of variance models may be used with repeated measures data, depending on the distribution of the multiple measurements. The two types of analyses are described and compared in Bock (1975 Chap. 7). (J. D. Finn) Full Reading ... PPT Download ...

Since the early 1970s path analysis and causal modeling have gained acceptance in educational research as well as in research in the social and behavioral sciences. The procedures employed have been developed to incorporate three main problems. First, in educational situations there are many outcomes as well as many explanatory factors to be considered. The analysis of the measures employed to represent these outcomes and explanatory constructs is confounded by problems of multicollinearity, measurement error, and validity. Measurements made on variables gain in strength and consistency if they are combined as related indicators of an underlying latent construct. Second, theory in educational research has advanced during the latter half of the twentieth century so that it is now possible to develop strong models that can be submitted to examination, and the estimation of the magnitude of the parameters of such models is of considerable theoretical interest and practical significance. Third, it is widely recognized that not only should the direct effects of explanatory variables be taken into consideration, but the mediating and moderating effects of such variables, as well as spurious and disturbance effects, should be examined. Those three issues have led to the development of latent variable path analysis and structural equation modeling (see Path Analysis and Linear Structural Relations Analysis).
Two general approaches have emerged in this field for the examination of models advanced from theoretical considerations. The first builds on the use of least squares regression analysis to predict and explain the effects of variables on one or more criteria. The emphasis in this approach is prediction and to maximize the amount of variance of the criteria explained by the predictors. The second approach builds upon maximum likelihood estimation procedures. This involves obtaining estimates of free parameters of a model, subject to specified constraints imposed by the fixed parameters, so that the covariance matrix derived from the estimations made is as close as possible to the covariance matrix based on the hypothesized model. Thus, the estimates obtained of the free parameters of the model are such that the difference between the covariance matrices of the observed data and the model are minimized. In this second approach the methods of parameter estimation employed distinguish between procedures that are dependent on the assumption of multivariate normality and those that are not. Normal theory estimation is associated primarily with the LISREL series of programs that employ both maximum likelihood estimation procedures and generalized and weighted least squares procedures (see Path Analysis
and Linear Structural Relations Analysis) and the work of Jöreskog and Sörbom (1989).
Asymptotic distribution free estimation is employed when the data are not multivariate normal and is associated with the work of Browne (1982, 1984) and Muthén (1984, 1987). This entry restricts itself to consideration of the approach that employs partial least squares (PLS) regression analysis to maximize prediction and the explanation of variance which was developed by Wold (1977, 1982). This approach is less well-known than the other approaches outside of continental Europe, with the developmental work having been carried out in Sweden and Germany. It has the clear advantages that no assumptions need be made about the shape and nature of the underlying distributions of the observed and latent variables. This permits the analysis of data of dichotomous variables that are not associated with an underlying continuous distribution, which is a distinct advantage for a variable such as sex of student, or for the use of variables to represent countries in cross-national comparative studies. Furthermore, the approach recognizes that nearly all data employed in educational research involve the use of complex cluster sample designs. As a consequence, procedures of statistical significance, that are heavily dependent on testing for statistical significance with assumptions of simple random sampling and multivariate normal distributions, are largely inappropriate. Least squares regression procedures are known from extensive experience to be robust. However, there is no proof, beyond very simple models which are equivalent to principal components analysis and canonical correlation analysis, that convergence in the iterative procedures employed is complete. There is the everpresent danger of a false minimum in the test for the iterative procedure, and thus an erroneous solution in the estimation process. Consequently, some form of testing by replication would appear to be essential to validate the solutions obtained with partial least squares analysis.
It should be noted, however, that partial least squares path analysis as a technique is quick in analysis, and convergence generally takes place rapidly; is flexible in use in the testing of complex models; and is relatively easy for a novice, but who has sound theoretical perspectives, to employ. Furthermore, while greater stability of the solution is attained with large samples, it does not demand large samples for effective operation, as is explained below. The maximum likelihood estimation approach is considered elsewhere (see Path Analysis and Linear Structural Relations Analysis). Full Reading ...

Of major interest and concern to the educational research worker is the observed variation of naturally occurring events in human behavior and educational practice. Not only does variation occur naturally in human characteristics, but it also arises in response to different treatment conditions acting to influence learning and the consequent stability and change in those characteristics. In addition, variation arises as a result of what has come to be known appropriately as “error,” which is associated with the random fluctuations of observations about an expected value. Such error occurs as a consequence of variability involving the observer, the variability in the procedures or the instruments used for observation, and variability in the object being measured.
Statistics has been referred to by Fisher (1970) as the study of the variation observed in the investigation of populations: The conception of statistics as the study of variation is the natural outcome of viewing the subject as the study of populations: for a population of individuals in all respects identical is completely described by a description of any one individual, together with the number in the group. The populations which are the object of statistical study always display variation in one or more respects. (p. 3)
Thus the educational research worker and statistician are necessarily concerned not only with the individual within the population under investigation, but also with the different conditions and circumstances that have contributed to the variation in the observations and measurements that are made. This entry presents the different measures of variation that are widely used in educational research. First, consideration is given to measures that relate to variation about the mean value. Second, a less extensive treatment is provided regarding measures that refer to variation about the median value. The former set of measures are of greater interest in practical investigations since they are rigorously defined, easily calculated, and more readily amenable to algebraic treatment and to systematic analysis. Thus, the mean and its associated measures of variation are generally employed in situations where the variables may be considered to involve interval or ratio data. However, the use of the median and its measures of variation is preferred in some situations, where the data are essentially ordinal in nature or where outlying values may distort the location of the mean and spread of values recorded. Since statistics is the study of variation, it is not surprising that reference should be made in other entries in this Handbook to the analysis of variance (see Variance and Covariance, Analysis of) and to the procedures by means of which the variation in a set of data can be partitioned into different components that can be ascribed to different factors (see Multilevel Analysis). These different factors may be associated with treatment conditions, naturally occurring variation, or different sources of error. (J.P. Keeves) Full Reading ... PPT Download ...

Discriminant analysis is the special case of regression analysis which is encountered when the dependent variable is nominal (i.e., a classification variable, sometimes called a taxonomic variable). In this case, either a single linear function of a set of measurements which best separates two groups is desired, or two or more linear functions which best separate three or more groups are desired. In the two-group application, the discriminant analysis is a special case of multiple regression, and in those applications where the criterion variable identifies memberships in three or more groups, the multiple discriminant analysis is a special case of canonical regression. However, because of its focus upon the parsimonious description of differences among groups in a measurement space, it is useful to develop the algebra of discriminant analysis separately from that of regression, and to have computer programs for discriminant analysis which are rather different in their printouts from general regression programs.

1. Aspects and History

Discriminant analysis has had its earliest and most widespread educational research applications in the areas of vocational and career development. Because education prepares people for a variety of positions in the occupational structures prevalent in their societies, an important class of educational research studies is concerned with the testing of theories about the causes of occupational placements and/or the estimation of prediction equations for allocating positions or anticipating such allocations. This research is characterized by criteria which are taxonomies of occupations or other placements, and predictors which are traits of the individuals who have been sampled from the cells of the taxonomy. Thus there are many independent variables which can be taken as approximately multivariate normal in distribution, and a dependent variable which is a nominal identifier of the cells of a taxonomy (or the populations in a universe). The resulting discriminant analysis design may be thought of as a reverse of the simple one-way “MANOVA” (multivariate analysis of variance) design.
Where MANOVA assumes a nominal independent variable and a multivariate normal dependent vector variable, discriminant analysis assumes a multivariate normal independent vector variable and a nominal dependent variable. Both methods share the assumption of equal measurement dispersions (variance–covariance structures) for the populations under study, so that much of the statistical inference theory of MANOVA is applicable to the discriminant design, especially the significance tests provided by Wilks' L, Pillai's V, Roy's q, and the Lawley–Hotelling U statistics (Tatsuoka, 1971 pp. 164–68, Timm, 1975 pp. 369–82). There is also a formal equivalence of discriminant analysis to the canonical correlation design with dummy variates representing group memberships (Tatsuoka, 1971 pp. 177–83). Discriminant analysis is also closely related to the statistical literature on the classification problem, since a discriminant space may be optimal for classification decisions in some situations (Rulon Ø et al. 1967 pp. 299–319, 339–41). However, a good discriminant program will report interpretive results which will not be obtained from a canonical correlation program, and modern computers so easily compute classifications in original measurement spaces of very large dimensionality that the reduction in dimensionality provided by a discriminant space no longer has great utility when the emphasis is on classification.
Discriminant analysis serves primarily to provide insight into how groups differ in an elaborate measurement space. It is most useful when the number of measurement variates is so large that it is difficult for the human mind to comprehend the differentiation of the groups described by a table of means. Usually it will be found that the major differences can be captured by projecting the group means onto a small number of best discriminant functions, so that an economical model of group differences is constructed as an aid to understanding. Often the discriminant functions will be theoretically interpretable as latent dimensions of the manifest variates. It is even possible to make a rotation of the discriminant dimensions to more interpretable locations, once the best discriminant hyperplane has been located. Thus the goals and procedures are not unlike those of factor analysis. (P. R. Lohnes) Full Reading ... PPT Download ...

During the 1970s and 1980s there was increasing acceptance and use of multivariate analytic procedures that examined the interrelations between the latent variables formed from defined sets of observed variates. Canonical variate analysis was first developed as early as 1935 by Hotelling (1935) and was largely unused because of the complexity of the computations that had to be carried out by hand. With the introduction of the electronic computer, programs which permit its use have become more readily available. However, the power of this analytical procedure is not generally acknowledged. Canonical variate analysis is in fact the general analytic method of which most other parametric statistical procedures, varying from t-tests and analysis of variance through to principal components analysis, factor analysis, and regression analysis and discriminant analysis are but special cases. Both historically and conceptually the basic analytical procedure employed in multivariate analysis is that of canonical variate analysis. Thus a sound understanding of the ideas and principles involved in canonical variate analysis provides an excellent foundation for the development of an understanding of other analytical procedures, such as partial least squares path analysis (PLS), linear structural relations (LISREL) analysis, and multivariate analysis of covariance (MANCOVA). This entry is concerned with the analytical procedure of canonical variate analysis. In this treatment, following Bartlett (1941), the term “variate” is used to refer to those observed measures that are introduced into the analysis, while the term “variable” is reserved for the latent constructs that are formed as a combination of the observed measures or variates. (J. P. Keeves, J. D. Thomson) Full Reading ... PPT Download ...

Scaling refers to a group of statistical methods that measure the relationships among psychological objects by arranging them in one or more dimensions. It is concerned with objects about which people can manifest some attitude or perception. Usually the experimenter wishes to know the relationship among a set of objects; that is, how far apart they are and in what relative directions they may lie. As a visual metaphor, the configuration reveals the mental organization of psychological objects. This entry discusses several of the
most useful scaling tasks and methods. The educational researcher should become aware of the extensive number and diversity of other scaling methods that exist by consulting the references.

1. Psychological Objects

Psychological objects can be tangible, such as chairs, stools and sofas, but they can also be non-physical, such as concepts, impressions, propositions, beliefs or tendencies. Psychological objects are most often presented as sentences or statements such as “There will always be wars” or “I hate war.” With young children, the objects are often pictures.

2. Judgments or Choices

The two major responses that subjects can make to a set of psychological objects are (a) similarity judgments and (b) choices (preferences). The type of response is determined by the measurement objectives of a particular study or experiment. If a psychological scale is to be constructed, then the responses to the objects should initially be judgments of similarity. Once the scale values have been created the preferences of a group of subjects can be determined. Figure 1 presents a diagrammatic outline for attitudinal measurement. First the psychological objects are chosen, dictated by the interests of the experimenter. Once the objects have been created or obtained they are presented in a task. If the task elicits judgment of similarities, the objects can be initially scaled. From such an analysis, a subset of them may be chosen and formulated into a more sensitive scaling instrument. The revised instrument can then be presented to the target group(s). Should preferences instead of judgments be obtained, a direct descriptive analysis is made. Such analyses can generate or test hypotheses. (P. Dunn-Rankin, Shuqiang Zhang) Full Reading ...

One of the most basic abilities of living creatures involves the grouping of similar objects, individuals, and so on, to produce a classification. Prehistoric people, for example, must have been able to recognize that many individual objects shared certain properties such as being edible, or poisonous, or ferocious, and so on. The idea of sorting similar things into categories is clearly a primitive one since classification, in the widest sense, is necessary for the development of language, which consists of words which help in the recognition and discussion of the different types of events, objects, and people encountered. Each noun in a language, for example, is a label used to describe a class of things which have striking features in common. So animals are named as cats, dogs, horses, and so on, and such a name collects individuals into groups. Naming is classifying.
As well as being a basic human conceptual activity, classification is also fundamental in most branches of science, since it involves two basic scientific functions: (a) the description of objects of interest or those under investigation, and (b) the establishment of general laws or theories by means of which particular events may be explained or predicted. Some areas where classification has played an important role are biology, where attempts at the classification of living organisms date from the time of Aristotle, and were a necessary prerequisite to the development of the evolutionary theories of Darwin; chemistry, where the classification of the elements in Mendeleyev's periodic table had a profound influence on uncovering the structure of the atom; and medicine, where a satisfactory classification of diseases is needed prior to investigating etiology and developing treatments. In education, researchers are often interested in producing classifications of both teachers and pupils. (B. S. Everitt) Full Reading ... PPT Download ...

Factor analysis is a technique for representing the relationships among a set of variables in terms of a smaller number of underlying hypothetical variables. It aims to describe the variation among a set of measures in terms of more basic explanatory constructs, and thus to provide a simpler and more easily grasped framework for understanding the network of relationships among those measures. Correlations might be computed, for example, among the scores of a group of students on measures of addition, subtraction, multiplication, division, vocabulary, and reading comprehension. A factor analysis of these correlations might show that the relationships among the tests could be almost completely explained in terms of two underlying variables, which might well be interpreted as computational ability and verbal ability.

1. Early Development of Factor Analysis

Although the technique of factor analysis is now applied in a wide variety of disciplines, it originated in the field of psychology. Toward the end of the nineteenth century, a number of psychologists turned their attention to experimental studies of intelligence and intellectual abilities. Spearman collected data to test his theory that mental activity could be explained in terms of a single central intellective function, “intelligence.” Finding high correlations between estimates of intelligence and students' scores on tests of weight, light, and pitch discrimination, he concluded that all branches of intellectual activity have in common one fundamental function (or group of functions), whereas the remaining or specific elements of the activity seem in every case to be wholly different from that in all others (Spearman 1904). Subsequently, in his two-factor theory, the fundamental function was described as a general factor, “g,” and the element specific to a particular activity as its specific factor, “s.” Spearman had noted that his matrices of correlations among intellectual abilities could be arranged hierarchically, showing a progressive decrease in value from left to right and from the upper to the lower rows of the table. He recognized that this would be the expected pattern of correlations if all mental processes reflected the operation of a single central intellective function, which operated at different levels of complexity. To test whether a set of correlations he had obtained among six variables conformed to this pattern, for instance, he computed the tetrad differences among the correlations, for example (r13r26 – r23r16). Finding that they were approximately zero, he confirmed the hypothesis that the correlations could be explained by one general factor.
The two-factor theory was challenged by Thomson and other psychologists on both theoretical and empirical grounds. Working with larger batteries of tests and larger numbers of cases, Burt identified verbal, numerical, and practical group factors in school subjects in addition to a general factor; a “group factor” is one which is represented only in certain similar types of tests but not in others. Spearman later admitted the necessity of group factors, and British factorists adopted a factor model which incorporated both a general factor and group factors. (D. Spearritt) Full Reading Full Download ...

Since the 1960s there have been major advances in educational research in the statistical procedures available for examining networks of causal relationships between observed variables and hypothesized latent variables. These procedures are considered in this entry using the related names of path analysis, causal modeling, and linear structural relations analysis. While the ideas underlying path analysis were first used in educational research by Burks (1928), these ideas lay largely dormant for over 40 years in this field until used by Peaker 1971) in England, largely because of the heavy computational work involved. The developments in education and the social and behavioral sciences have followed those advanced earlier in the field of genetics (Wright Ø 1934). The techniques were introduced more generally into the social sciences by Duncan (1966) and articles by Land (1969), Heise (1969), and Duncan (1969) served to systematize the procedures being used and to clarify many of the issues involved. The approach employing these techniques enables the

investigator to shift from verbal statements of a complex set of interrelationships between variables to more precise mathematical ones, which are commonly represented in diagrammatic form, and to estimate the magnitudes of the causal relationships involved. Stokes (1968, 1974) developed the procedures of path analysis from first principles, which emerged as equivalent to ordinary least squares regression analysis (see Keeves, 1988). A more general term for the analytical techniques involved, namely“structural equation modeling,” is also used. As implied by this name a set of mathematical equations is employed to formulate the structural relations which are hypothesized to exist between a network of observed and latent variables, and the model is tested to determine the extent to which it

accounts for the covariation between the observed measures. A variety of strategies is available for the examination of such models, and this entry considers the different approaches that are now widely employed in educational, behavioral, and social science research in the field of structural equation modeling. This entry makes a distinction between path analysis, in which least squares regression procedures are used to maximize the variance explained, and procedures in which the model is estimated and the fit between the model and the data examined. The former is considered in greater depth (see Path Analysis with Latent Variables) and the latter in the discussion that is presented below.

It is rare in educational research for an investigator to proceed with an inquiry in which evidence is assembled without maintaining some theoretical perspectives, whether implicitly held or explicitly stated, that guide the design of the investigation and that influence the analyses carried out. In such analyses there is interest in examining the patterns in the covariation between measures and testing whether these patterns are consistent with the theoretical perspectives which are held. However, it is not possible to impute causal relationships from a study of the covariation between measures, and it is necessary to assume a scheme of causation derived from the theoretical perspectives which are maintained. Where these theoretical perspectives extend beyond simple description, they commonly involve ideas of causation which are probabilistic and stochastic in nature in so far as a time sequence is also involved. It is these ideas of causation that are examined in the analysis of data. (A. C. Tuijnman, J. P. Keeves) Full Reading ... PPT Download...

Prediction represents an effort to describe what will be found concerning an event or outcome not yet observed on the basis of information considered to be relevant to the event. Typically, there is a temporal dimension to prediction when, say, ability test scores are used to forecast future achievement in a course of study.

1. Predictor and Criterion Variables

The information that is used to make a prediction is typically referred to as a predictor. In any prediction study there is at least one predictor variable. Predictor variables can be either quantitative—for example, scores on a test—or qualitative—for example, type of course in which enrolled. It is possible to combine qualitative and quantitative variables in a prediction study. The event or outcome to be predicted is typically referred to as a criterion. There are several types of criterion variables. One of the most common is performance on some quantitative continuous variable such as an achievement test or a grade point average. Other criterion variables could be qualitative in nature. When a counselor helps a student make a choice of career or course of study, the counselor is at least implicitly making a prediction about a qualitative variable, career choice, or course of study, in which the student is likely to succeed and from which he or she can derive satisfaction. The criterion in this case can be regarded as membership in a particular group. In more complex cases, a criterion may be multidimensional in character, as when one is interested in predicting an individual's profile on a number of variables such as a battery of achievement measures. While possible, simultaneous prediction on a number of criterion variables is quite rare. Prediction studies can be highly varied depending on the nature of the population under study and the number and types of predictor and criterion variables used. However, there are a number of common elements in prediction studies. The first element involves identifying the outcome or event to be predicted. Education, social, or business necessity is usually the basis for such a choice. The second element is to develop or select a measure that will serve as a criterion variable. This crucial step of the process generally receives far less attention than it should. If one decides to use “success in college” as a criterion in a prediction study, then how is success to be defined and measured? Obviously, there are many ways to do this task. Unfortunately, there is no clearly correct way to define and measure success. Any definition of success will have its limitations, and any measure based on a particular dimension will be somewhat deficient. The use of an earned grade point average in college as a measure of success, for example, will only reflect the standards used by a particular group of instructors in a set of courses and will ignore performance in noninstructional aspects of college life. Furthermore, the standards used by a particular set of instructors may not reflect the standards of the institution as a whole. In fact, it is possible that there may be no uniform agreement on a set of standards that will serve to define success across all the instructional areas of the institution. Thus, even a widely used criterion variable such as the grade point average suffers from a number of weaknesses. (R. M. Wolf, W. B. Michael) Full Reading ... PPT Downlod ...

In its most general sense, the term “correlation” refers to a measure of the degree of association between two variables. Two variables are said to be correlated when certain values of one variable tend to co-occur with particular values of the other. This tendency can range from nonexistent (all values of variable 2(Y) are equally likely to occur with each value of variable 1(X)), to absolute, (only a single value of Y occurs for a given value of X). A correlation coefficient is a numerical index of the strength of this tendency. This entry is concerned with the properties and uses of correlation coefficients. There are many special types of correlation coefficients. All share the characteristic of being measures of association between two variables. The difference lies in how the variables are defined. What is commonly called the “correlation coefficient” (or, more properly, the “Pearson product–moment correlation coefficient” or PMC) is an index of association between two variables where both variables are considered as single observed variables. The association between scores on a reading readiness test and a later measure of reading achievement is an example of such a correlation. Although each measure is the sum of scores on several items or subtests and represents a complex human characteristic, it is treated as a single undifferentiable variable for the purposes of analysis.

At the other end of the spectrum is a procedure called set correlation (Cohen, 1982). Like the ordinary PMC, it is a measure of the association between two variables; however, in this case the association is between two unobservable variables that are weighted additive combinations of multiple measured variables. The unobserved variables are often called latent variables or composite variables. If, for example, the reading readiness and achievement measures each yielded several subscales, the correlation between a weighted combination of one set of subscales and a weighted combination of the other would be a canonical correlation. Set correlation is a generalization of canonical correlation in which additional sets of variables can be statistically removed from one or both of the sets being correlated (see partial correlation below). Full Reading ...

When you open SPSS Program, it displays SPSS Opening Screen. From the this screen, SPSS asks whether you want to Run the tutorial, Type in data, Run an existing query, Create new query using Database Wizard or Open an existing data source.

At the moment, we want to talk about how to open an existing data file.

If you want to open an existing data file (e.g. one of the files accompanied SPSS tutorials), click on the Opening an existing data source button from the opening screen, then on More Files. This will take you to search through the various directories on your computer to find where your data file is located. Find the file you want to use and double click on it or click on Open from your Menu. The data file will open in front of you with the Data Editor label.



Posted to you by SPSS Books

There are many books on SPSS available in the market and it is not easy for everyone to decide which one is the most appropriate for him/her to use or to buy. One of the best ways to help you make decision by reading users' reviews and rating results. Here, we present a page to help you to make decision when you need to buy SPSS guide or manual books. Find more ...

"I have been studying for GMAT for a few weeks now and I think the OG is a great source for a good practice on real GMAT questions. But having said that, use the OG only once you grasp all the basics and when you are almost ready to take the GMAT. The OG does not contain any math conceptual content, so use it only for practice towards the end." See the book ...

"In order to get a real GMAT prep - I would recommend you buy the set of 9 books of "EZ Solutions," or whichever book(s) you think you need most help with. These books are very detailed oriented and cover everything on the GMAT math. First use the review modules to get the concepts and then use the workbooks for practice. Some of the most challenging topics in GMAT math, such as permutation/combination, probability, geometry problems, complex word problems, etc., are brilliantly explained in these books. Also, the EZ Advanced Workbook is a "must have" if you are aiming for a high score. It may be a good idea to buy Barron's without the CD (don't waste money on the CD version for any book). After doing all this, use the OG and you will see a dramatic difference in your scores. If you still need more practice, you may consider buying the Kaplan book (but really not needed). Save your money by not buying anything else. You don't even have to take any of those pricy courses. I followed this process and my scores have jumped from the 500-range to the 700-range. Good luck!" See the book ...

With SPSS 16.0 you can get several data files open at one time and simply choose from the menu Window to activate the data file you want to work with. It's just like what you do when you work with Microsoft Word Processors.

So, SPSS 16.0 can display several data files at the same time.


Another thing new, SPSS 16.0 has a facility to help you to keep records of what you have done with your works. When you start opening a new data file, SPSS 16.0 by default will open a new window called SPSS Viewer and make a record of the data files you open.

Technorati Tags: SPSS, SPSS 16, SPSS Manual, SPSS Tutorial

Evaluation studies often lack sophistication in their statistical analyses, particularly where there are small data sets or missing data. Until recently, the methods used for analysing incomplete data focused on removing the missing values, either by deleting records with incomplete information or by substituting the missing values with estimated mean scores. These methods, though simple to implement, are problematic. However, recent advances in theoretical and computational statistics have led to more flexible techniques with sound statistical bases. These procedures involve multiple imputation (MI), a technique in which the missing values are replaced by m > 1 estimated values, where m is typically small (e.g. 3–10). Each of the resultant m data sets is then analysed by standard methods, and the results are combined to produce estimates and confidence intervals that incorporate missing data uncertainty. This paper reviews the key ideas of multiple imputation, discusses the currently available software programs relevant to evaluation studies, and demonstrates their use with data from a study of the adoption and implementation of information technology in Bali, Indonesia.
Introduction Missing observations occur in many areas of research and evaluation (Kline 1998). When data are collected by surveys, questionnaire responses may be incomplete because some respondents refuse to answer certain questions. In longitudinal studies, subjects may drop out early or be unavailable during one or more data collection periods. These types of missing data are unintended and uncontrolled by the researcher but the overall result is that useful data collected from a survey cannot be analysed in detail because of the extent of missing data. This paper introduces and discusses the three ad hoc methods for dealing with missing data and then focuses on the software developed to process missing data by the multiple imputation method. (I.G. Darmawan) Full Reading ... Full Download ...

In its most general sense, the term “correlation” refers to a measure of the degree of association between two variables. Two variables are said to be correlated when certain values of one variable tend to co-occur with particular values of the other. This tendency can range from nonexistent (all values of variable 2(Y) are equally likely to occur with each value of variable 1(X)), to absolute, (only a single value of Y occurs for a given value of X). A correlation coefficient is a numerical index of the strength of this tendency. This entry is concerned with the properties and uses of correlation coefficients. There are many special types of correlation coefficients. All share the characteristic of being measures of association between two variables. The difference lies in how the variables are defined. What is commonly called the “correlation coefficient” (or, more properly, the “Pearson product–moment correlation coefficient” or PMC) is an index of association between two variables where both variables are considered as single observed variables. The association between scores on a reading readiness test and a later measure of reading achievement is an example of such a correlation. Although each measure is the sum of scores on several items or subtests and represents a complex human characteristic, it is treated as a single undifferentiable variable for the purposes of analysis.
At the other end of the spectrum is a procedure called set correlation (Cohen, 1982). Like the ordinary PMC, it is a measure of the association between two variables; however, in this case the association is between two unobservable variables that are weighted additive combinations of multiple measured variables. The unobserved variables are often called latent variables or composite variables. If, for example, the reading readiness and achievement measures each yielded several subscales, the correlation between a weighted combination of one set of subscales and a weighted combination of the other would be a canonical correlation. Set correlation is a generalization of canonical correlation in which additional sets of variables can be statistically removed from one or both of the sets being correlated (see partial correlation below). Full Reading ... Full Download ... PPT Download ...