Reading Statistics and Research

Bi-variate correlation assumptions not met

1. Identifying Outliers 2. Trimmed Samples 3. Winsorized Samples 4. Logarithmic Samples

Mixed ANOVA Assumptions

1. Independence between groups and independence within groups. 2. Normality. 3. Homogeneity of Variance. 4. Sphercity.

Multiple Regression Assumptions

1. Independence of observations 2. Criterion is normally distributed, aka normality in the arrays/residuals. 3. Linear relationship between X's and Y. 4. Homoscedasticity.

One way within subjects ANOVA assumptions

1. Independence of observations 2. Normality 3. Sphericty

Mixed ANOVA Null Hypothesis of Main Effects

1. Main effect for the within subjects factor: In the population, all levels of the factor have equal marginal means. 2. Main effect for between-groups factor: In the population, all levels of the factor have equal marginal means

Mixed ANOVA Alternative Hypothesis of Main Effects

1. Main effect for the within subjects factor: In the population, there is some inequality between the factors levels marginal means 2. Main effect for between groups factor: In the population, there is some inequality between the factors levels marginal means

One-way between groups ANOVA assumptions

1. Normality 2. Homogeneity of variance 3. Independence between and within groups

Chi-square test of association assumptions

1. Normality 2x2 tables less than 5 or larger than 2x2 not more than 20% less than 5. 2. Inclusion of nonoccurances. 3. Independence within/between groups.

Bi-variate correlation assumptions

1. Univariate Normality 2. Bivariate Normality 3. Linearity 4. Homescedasticity 5. Independence of Observations

Linear regression assumptions

1. Univariate Normality 2. Bivariate Normality 3. Linearity 4. Homescedasticity 5. Independence of Observations

Confidence Intervals (Interval Estimation)

A confidence interval (CI) is simply a finite interval of score values on the dependent variable. Such an interval is constructed by adding a specific amount to the computed statistic (thereby obtaining the upper limit of the interval) and by subtracting a specific amount from the statistic (thereby obtaining the lower limit of the interval

Bi-variate correlation purpose

A correlation that assesses the relationship or association of two variables (x and y). The Pearson correlation coefficient (r) is a parametric statistic that provides information about the direction and magnitiude of assocation between two variables.

Mixed ANOVA Purpose

A mixed model ANOVA includes one or more between groups factors in addtion to one or more within subjects factors

The Cochran Q Test

A test developed by Cochran is appropriate for the situation where the researcher wishes to compare three or more related samples with respect to a dichotomous dependent variable

Two-Way ANOVA

A two-way ANOVA always involves two independent variables. Each independent variable,or factor,is made up of,or defined by,two or more elements called levels. When looked at simultaneously, the levels of the first factor and the levels of the second factor create the conditions of the study to be compared. Each of these conditions is referred to as a cell.

Quasi-experimental

Always has manipulation, some control in most cases, usually lacks random assignment, weak causal inferences may be drawn, moderate internal validity and moderate external validity

One-way between groups ANOVA post-hoc tests

Bonferonni. The significant difference between these groups means of the rhyming group and the visual group (p=.016). Looking at the descriptive statistics we can conclude that in the population the mean on number of words recalled is higher for the visual group than for the rhyming groups.

Reliability

Can be summed up by the word consistency

Dependent Variable (Regression)

Criterion variable, outcome variable, response variable

Reliability Coefficient

Descriptive summary of the data's consistency normally assumes a value between 0.00 and +1.00

Mixed ANOVA Results Interaction Effect

F = 1.245, df = 5.371, 64.452, p = .297. The result is not significant. We do not reject the null hypothesis and conlude that in the population, the effect of each factor is consistent across the levels of the other factor. In other words there is no interaction effect in the population. Partial eta sqaured = .784, this tells us that 78.4% of the variance in DV that is not accounted for by other effects is accounted for by or associated with the time in this sample.

Mixed ANOVA Results Between Groups Main Effect

F = 1.644, df = 3.36, p = .196, The result is not significant. We do not reject the null hypothesis and conclude that, in the population, all levels of the factor type of therapy have equal marginal means

Mixed ANOVA Results Within Subjects Main Effect

F = 130.33, df = 1.790, p = .000. The result is significant. We can reject the null hypothesis and conclude that, in the population, there is some inequality between the marginal means of the levels of time.

One-way between groups ANOVA results

F = 3.840, df = 4,45; p = .009. The result is significant, Therefore we can reject the null hypothesis and conclude that there is at least one difference among the groups means in the population. With one factor, eta squared = partial eta squared. Eta squared (.254) tells us that 25.4% of the variance in this sample for the DV is shared with or related to the IV.

One way within subjects ANOVA results

F = 55.897, df = 4,32 p < .001. The result is significant. We can reject the null hypothesis and conclude that there is some inequality between the groups. In order to discuss the importance of these results, refer to eta-sqaured = .875. This tells us that 87.5% of the variance in the DV is accounted for by or associated witht the IV in this sample

Measurement Model

First, it posits the existence of the study's latent variables. Second, it asserts that these latent variables manifest themselves in the study's observed variables

MANOVA

For any of the tests on means considered in earlier chapters, a multivariate procedure exists to handle the case of multiple dependent variables. Stated differently,for any ANOVA that has been designed for data on a single dependent variable, there is a

Non-parametric Test for Repeated Measures One Way ANOVA

Friedman's test

Three Null Hypothesis of a Two Way ANOVA

H01:There is no significant difference in the test anxiety level of groups subjected to REBT therapy and control after treatment. H02:There is no significant difference in the test anxiety level of groups with moderate and high entry anxiety level at the end of treatment. H03:There is no significant interaction effect of treatment by entry test anxiety level on test anxiety level at the end of treatment

Bi-variate correlation null hypothesis

H0: p = .00. The population correlation is equal to .00

Linear regression null hypothesis

H0: population beta = .00, where beta is the standardized regresson coefficient, for use in predicting Z scores on y from z scores on x. In the population, the criterion cannot be predicted from the predictor.

Bi-variate correlation alternative hypothesis

H1: p ≠ .00. The population correlation is not equal to .00 H1: p > .00 or p < .00 (one tailed) The population correlation is greater/less than .00

Linear regression alternative hypothesis

H1: population beta ≠ .00 (two-tailed). In the population, the predictor significantly predicts the criterion H1: population beta > .00 or population beta < .00 (one tailed)

Quota Samples (Non Probable)

Here,the researcher decides that the sample should contain Xpercent of a certain kind of person (or object), Y percent of a different kind of person (or object), and so on. Then, the researcher simply continues to hunt for enough people/things to measure within each category until all predetermined sample slots have been filled

Mann-Whitney U test assumptions

Homogeneity of variance

Power Analysis

How large should the sample be? At this point (and also in Steps 7 through 9), the researcher functions like a robot who performs a task, referred to as a

Concurrent Validity

If the new test is administered at about the same time that data are collected on the criterion variable, then the term

One way within subjects ANOVA results

If you obtain a significant F statistic for the within subjects ANOVA, SPSS allows you to run two post-hoc tests: Bonferonni or Sidak. Either approach can be used but Sidak is more conservative than the Bonferrroni.

Predictive Validity

If, however, people were given the new test years before they took the criterion test,then rwould be a measure of

Regression (Differences from correlation)

In a analysis involving A and B one of the two variables needs to be identified as the dependent variable and the other variable must be seen as the independent variable

Test-Retest Reliability

In many studies, a researcher measures a single group of people (or animals or things) twice with the same measuring instrument, with the two testings separated by a period of time

Attrition (nonresponse bias)

In many studies, less than 100 percent of the participants remain in the study from beginning to end. In some instances, this problem arises because the procedures or data-collection activities of the investigation are aversive, boring, or costly to the participan

Purposive Samples (Non Probable)

In some studies, the researcher starts with a large group of potential participants. To be included in the sample, however, members of this large group must meet certain criteria established by the researcher because of the nature of the questions to be answered by the investigation.

Convenience Sample (Non-Probable)

In some studies,no special screening criteria are set up by the researchers to make certain that the individuals in the sample possess certain characteristics. Instead, the investigator simply collects data from whoever is available or can be recruited to participate in the study

Refusal to Participate (nonresponse bias)

In studies where individuals are asked to participate, some people may decline.

Carryover Effects

In studies where the repeated-measures factor is related to different kinds of treatments, the influence of one treatment might interfere with the assessment of how the next treatment works.

One-way within subjects ANOVA null hypothesis

In the population, all levels of the factor have equal means.

One-way between groups ANOVA null hypothesis

In the population, all of the groups have equal means. H0: m1 = m2 = m3

Mixed ANOVA Null Hypothesis of Interaction Effect

In the population, the effect of each factor is consistent across the levels of the the other factor

Mixed ANOVA Alternative Hypothesis of Interaction Effect

In the population, the effect of each factor is no consistent across the levels of the other factor

Chi-square test of association alternative hypothesis

In the population, there IS an association between variables. An association means that the proportational frequency distribution on each variable is INCONSISTENT across the levels of the other variable.

Chi-square test of association null hypothesis

In the population, there is no association between the variables. No association means the proportional frequency distribution on each variable is CONSISTENT across the levels of the other variable

One way within subjects ANOVA alternative hypothesis

In the population, there is some inequality between the means of the factor being tested

Multiple Regression Null Hypothesis

In the population, this set of IV's taken together does not predict or relate to scores on Y. Population R squred = 0

Multiple Regression Alternative Hypothesis

In the population, this set of IV's taken together predicts or relates to scores on Y. Population R squared > 0.

Wilcoxon rank sum test assumptions

Independence between and within

Kruskal-Wallis One Way ANOVA by ranks

Independence between and within, homogeneity of variance

Friedman's test assumptions

Independence within

Wilcoxon matched pairs signed ranks test assumptions

Independence within

Alternate-Forms Reliability

Instead of assessing stability over time, researchers sometimes measure people with two forms of the same instrument.The two forms are similar in that they supposedly focus on the same characteristic (e.g., intelligence) of the people being measured, but they differ with respect to the precise questions included within each form

Internal Consistency Reliability

Instead of focusing on stability across time or on equivalence across forms, researchers sometimes assess the degree to which their measuring instruments possess internal consistency

One-way Between Groups ANOVA

Kruskal-Wallis One-Way ANOVA by ranks

Quasi-experimental

Manipulation - Always Random Assignment - Rarely Control - Sometimes Causal Inference - Weak

Experimental

Manipulation - Always Randon Assignment - Always Control - Always Causal Inference - Strong

Non-experimental

Manipulation - Never Random Assignment - Sometimes Control - Rarely Causal Inference - None

True-experimental

Manipulation, random assignment, control, strong causal inferences may be drawn, high internal validity but poor external validity

Construct Validity

Many measuring instruments are developed to reveal how much of a personality or psychological construct is possessed by the examinees to whom the instrument is administered

Non-experimental

Non-causal, observes events as they occur, correlational (no manipulation), poor internal validity, high external validity

Assumptions of one way ANOVA

Normality, homogeneity of variace

Split-half reliability coefficent

One of the procedures that can be used to obtain the internal consistency reliability coefficient involves splitting each examinee's performance into two halves, usually by determining how the examinee did on the odd-numbered items grouped together (i.e., one half of the test) and the even-numbered items grouped together (i.e., the other half)

Independent Variable (Regression)

Predictor variable, explanatory variable

Multiple Regression Results

R = .753, R squared = .567, adjusted R squared = .557, F = 55.942, df = 3, 128, p < .001. Results are significant. Therefore, we can reject the null and conclude that population R squared > 0; this set of IV's taken together predicts scores on the DV in the population

Multiple Regression R squared Results

R squared = .567, therefore in this sample father, mother, and sister passivity scores taken together, explain 56.7% of the variance in boys passivity scores. Adjusted R squared = .557, so we estimate that in the population, 56% of the variance of the DV is associated with these IV's taken together.

Latent Variables

Regardless of how many factors are predicted to exist, each one is often referred to as a

Independent Samples Chi Square Test

Researchers frequently wish to compare two or more samples on a response variable that is categorical in nature. Because the response variable can be made up of two or more categorie

The Wilcoxon Matched Pairs Signed Ranks Test

Researchers frequently wish to compare two related samples of data generated by measuring the same people twice (e.g., in a pre/post sense) or by measuring two groups of matched individuals just once. If the data are interval or ratio in nature and if the relevant underlying assumptions are met,the researcher will probably utilize a correlated t-test to compare the two samples. On occasion, however, that kind of parametric test cannot be used because the data are ordinal or because the t-test assumptions are untenable (or considered by the researcher to be a nuisance). In such situations, the two related samples are likely to be compared using the

Criterion-Related Validiity

Researchers sometimes assess the degree to which their new instruments provide accurate measurements by comparing scores from the new instrument with scores on a relevant criterion variable. The new instrument under investigation might be a short, easy-to-give intelligence test, and in this case the criterion would probably be an existing reputable intelligence test (possibly the Stanford-Binet)

Interrater Reliability

Researchers sometimes collect data by having raters evaluate a set of objects,pictures, applicants, or whatever. To quantify the degree of consistency among the raters, the researcher computes an index o

Standard error of measurement (SEM)

Some researchers, when discussing reliability, present a numerical value for the standard error of measurement (SEM) that can be used to estimate the range within which a score would likely fall if a given measured object were to be remeasured

Friedman's Two Way Analysis of Variance of Ranks

The Friedman test is like the Wilcoxon test in that both procedures were developed for use with related samples. The primary difference between the Wilcoxon and Friedman tests is that the former test can accommodate just two related samples whereas the Friedman test can be used with two or more such samples

Standard Error (Interval Estimation)

The distribution of sample statistics alluded to in the preceding paragraph is called a sampling distribution, and the standard deviation of the values that make up such a distribution is called a

Linear regression results

The result is significant, r = .637, r2 = .406, p < 001. We can reject the null hypothesis, population beta = .00, and state that, in the population, we can predict self-esteem scores from GPA scores. Based on our results, we can state that in the sample, 40.6% of the variance in self-esteem is associated with GPA scores. Furthermore, the estimate for the population suggests tht 38.4% of the variance in self-esteem is associated with GPA scores.

Bi-variate correlation results

The result is significant, r = .984, p < .001. We can reject the null hypothesis and state that in the population, there is a strong positive relationship between stress and health. 97% of the variance in stress is associated with health.

Logistic regression results (overall)

The results are significant, Model chi-square = 28.24, df = 2, p < .001. We reject the null hypothesis and conclude that in the population this set of predictors improve prediction over the null model. Given that Nagelkerke R squared = .615, it suggests a strong degree of association between the model and group membership. We would also observe that the model significantly improved classification sucess 23.9% over and above the null model

Post Hoc Analysis

The three most frequently used procedures are called the Bonferroni test, the Tukey test, and Scheffé test.

Validity Coefficient

The validity of either of those new tests would be determined by (1) finding out how various people perform on the new test and on the criterion variable, and (2) correlating these two sets of scores. The resulting r is called the

Multiple Regression

This form of regression involves, like bivariate regression, a single dependent variable. In multiple regression, however, there are two or more independent variables. Stated differently,multiple regression involves just one Yvariable but two, three, or more X variables.

Bonferroni technique

This is usually accomplished by simply dividing the desired Type I error risk for the full study by the number of times the hypothesis testing procedure is going to be used

Kuder-Richardson #20

This procedure, like the split-half procedure, uses data from a single test that has been given just once to a single group of respondents. After the full test is scored, the researcher simply puts the obtained data into a formula that provides the K-R 20 reliability coefficient. The result is somewhat like a split-half reliability, but better, because the split-half approach to assessing internal consistency yields a result that can vary depending on which items are put in the odd-numbered slots and which are placed in the even-numbered slots

Cronbach's alpha

This technique is identical to K-R 20 whenever the instrument's items are scored in a dichotomous fashion (e.g., "1" for correct, "0" for incorrect). However, alpha is more versatile because it can be used with instruments made up of items that can be scored with three or more possible value

Linear regression purpose

To allow prediction of Y on the basis of knowledge about X

Coefficient of equivalence

To quantify the degree of alternate-forms reliability that exists, the researcher administers two forms of the same instrument to a single group of individuals with a short time interval between the two testings.3 After a score becomes available for each person on each form, the two sets of data are correlated. The resulting correlation coefficient is interpreted directly as the alternate-forms reliability coefficient.4 Many researchers refer to this two-digit value as the coefficient of equivalence.

Interval Estimation

To understand how interval estimation works, you must become familiar with three concepts: sampling errors, standard errors, and confidence intervals (CIs)

Regression (Differences from correlation)

Tries to accomplish one or the other two goals: prediction or explanation

Chi-square test of association purpose

Use this statistic when there are two categorical variables, when we have indepdendent samples and we want to know if peoples status on one of the categorical variables is associated with their status on another

Sampling Error (Interval Estimation)

When a sample is extracted from a population, it is conceivable that the value of the computed statistic will be identical to the unknown value of the population parameter. Although such a result is possible, it is far more likely that the statistic will turn out to be different from the parameter. The term sampling error refers to the magnitude of this difference.

Purpose of a One Way ANOVA

When a study has been conducted in which the focus is centered on three or more groups of scores, a one-way ANOVA permits the researcher to use the data in the samples for the purpose of making a single inferential statement concerning the means of the study's population

Kruskal=Wallis H test

When researchers wish to compare three or more such groups, they more often than not utilize the Kruskal-Wallis H test

Cluster Samples (Probable)

When this technique is used to extract a sample from a population,the researcher first develops a list of the clusters in the population. The clusters might be households,schools,litters,car dealerships, or any other groupings of the things that make up the population. Next, a sample of these clusters is randomly selected

Non-parametric Test for Dependent Samples t-test

Wilcoxon matched pairs signed ranks test

Non-parametric Test for Independent Samples t-test

Wilcoxon rank sum test or Mann-Whitney U test

Test-retest reliability coefficient

With a test-retest approach to reliability, the resulting coefficient addresses the issue of consistency, or stability, over time. For this reason, the test-retest reliability coefficient is frequently referred to as the coefficient of stability. As with other forms of reliability, coefficients of stability reflect high reliability to the extent that they are close to 1.0

Content Validity

With certain tests, questionnaires, or inventories, an important question concerns the degree to which the various items collectively cover the material that the instrument is supposed to cover. This question can be translated into a concern over the instrument's. Usually, an instrument's standing with respect to content validity is determined simply by having experts carefully compare the content of the test against a syllabus or outline that specifies the instrument's claimed domain. Subjective opinion from such experts establishes—or does not establish—the content validity of the instrument

Regression (Differences from correlation)

With correlation, there is just one thing that can be focused on: the sample correlation coefficient. With regression, however, inferences focus on the correlation coefficient, the regression coefficient(s), the intercept,the change in the regression coefficient,and something called the odds ratio

One Sample Chi Square Test

With this kind of chi-square test, the various categories of the nominal variable of interest are first set up and considered. Second, a null hypothesis is formulated. The for the one-sample chi-square test His simply a specification of what percentage of the population being considered falls into each category. Next, the researcher determines what percentage of the sample falls into each of the established categories. Finally, the hypothesis testing procedure is used to determine whether the discrepancy between the set of sample percentages and those specified in is large enough to permit to be rejected

Active Factor

a participant's status on the factor is determined within the investigation, because active factors deal with conditions of the study that are under the control of the researcher. Simply put, this means that the researcher can decide, for any participant, which level of the factor that participant will experience

Multiple Regression Purpose

a. to predict scores on a DV from scores on a set of IV's b. to study the degree of association between the one DV and a set of IV's

Two Way Mixed ANOVA

always involves one between-subjects factor and one within-subjects factor, the number of levels in each factor vary from study to stud

Stepwise Multiple Regression

analysis is analogous to the process of preparing a soup in which the ingredients are tossed into the pot based on the amount of each ingredient. Here the stock goes in first (because there is more of that than anything else), followed by the vegetables, the meat, and finally the seasoning

Exongenous Variables

are considered to be the independent variables in an SEM model,because they are thought to be on the front end of a causal relationship

Endogenous Variables

are considered to be the model's dependent variables, because they are believed to be affected by one or more of the other latent variables.

Observed Variables

are individual items in a questionnaire or survey; however, such variables can be anything the researcher thinks is a good proxy for, or representative of, the underlying hypothesized latent variable(s).

Latent Variables (SEM)

are traits or constructs that cannot be observed or measured directly. Examples of such variables would be your level of test anxiety, the amount of intelligence you possess, your trustworthiness, your fear of spiders, how much you enjoy hiking in the woods, and so on

Logistic Regression

as in any bivariate or multiple regression, there is one and only one dependent variable. Here, however, the dependent variable is categorical. Although the dependent variable can have three or more categories, thus making the logistic regression multinomial in nature, we consider here only situations where the dependent variable is dichotomous at least one independent variable is involved in any logistic regression. Almost always, two or more such variables are involved. As in multiple regression, these variables can be either quantitative or qualitative in nature

Assigned Factor

deals with a characteristic of the things being studied that they "bring with them" to the investigation. In situations where the study focuses on people, for example, such a factor might be gender, handedness, birth order, intellectual capability, color preference, grade point average (GPA), or personality type.

F Value (one way ANOVA)

divided the mean (MS) on the between groups row by the MS on the within groups row

MS Value (one way ANOVA)

dividing the rows sum of squares (SS) value by its df value

Multiple Regression Beta Weights

each beta weight reflects each IV's independent contribution to predicting the DV, therefore a beta weight is akin to but not identical to a semi-partial correlation.

Structural Model

extends beyond the measurement model and posits the way in which the latent variables are related to each other. This model stipulates which pairs of latent variables have a causal connection, which pairs of variables are related but not in a causal manner, and which pairs of variables are independent of each other

Response Rate

has been coined to indicate the percentage of sample individuals who supply the researcher with the requested information

one-way ANOVA

has one independent variable, it focuses on one dependent variable, and it involves samples that are independent

Exploratory Factor Analysis

he researcher has little or no idea as to number or nature of factors that will emerge from the analysis. With this kind of factory analysis, it is as if the researcher is about to visit a new art museum that has just been built and filled with artistic treasures. Once inside, the researcher discovers, for the first time, how to navigate through the different rooms, where the different installations are located, and what specific items of art have the strongest personal appeal

Logistic regression alternative hypothesis

in the population, this set of predictors does improve prediction over the null model

Logistic regression null hypothesis

in the population, this set of predictors does not improve prediction over the null model

Logistic regression assumptions

independence of errors, linearity, no specification errors

Bivariate Regression

involves just two variables and is used most frequently to see how well scores on the dependent variable can be predicted from data on the independent variable.

non-pairwise comparisons

involves three or more groups, with these comparison groups divided into two subsets

Factor Analysis

is a procedure that attempts to reduce the complexity of a multi-variable data set so it becomes easier for people to use the data in applied settings or in the development/refinement of theory. In a word,the goal of factor analysis is parsimony. The main question is simple: Can the people, animals, or things that have been measured on several variables be described, accurately, by means of a small number of numerical descriptors rather than by scores on each of the initial variables?

one-way ANOVA null hypothesis

is always set up to say that the mean score on the dependent variable is the same in each of the populations associated with the study. The null hypothesis is usually written by putting equal signs between a set of s, with each representing the mean score within one of the populations. For example, if there were four comparison groups in the study, the null hypothesis would be .

Residual Error

is associated with latent variables, but only latent variables that function as dependent variables. This kind of error can be thought of as what is left after the relevant independent (exogenous) variable(s) explain, or account for, as much variability in the dependent (endogenous) variable as it or they can

Validity

is captured nicely by the word accuracy

Systematic Sample (Probable)

is created when the researcher goes through an ordered list of members of the population and selects,for example,every fifth entry on the list to be in the sample. (Of course, the desired size of the sample and the number of entries on the list determine how many entries are skipped following the selection of each entry to be in the sample.) So long as the starting position on the list is determined randomly, each entry on the full list has an equal chance of ending up in the sample

Measurement Error

is created whenever data are gathered by means of a measuring instrument or process that has less than perfect reliability. Because perfectly reliable measuring instruments are few and far between, measurement error is almost always embedded in the scores created when researchers try to tap into observed variable

The Median Test

is designed for use when a researcher wishes to compare two or more independent samples. If two such groups are compared, the median test is a nonparametric analog to the independent-samples t-test. With three or more groups, it is the nonparametric analog to a one-way ANOVA

Random assignment

is how you assign the sample that you draw to different groups or treatments in your study (randomly to treatment groups)

Random selection

is how you draw the sample of people for your study from a population (randomly from the population of interest)

The Mann-Whitney U Test

is like the two-sample version of the median test in that both tests allow a researcher to compare two independent samples. Although these two procedures are similar in that they are both considered to be nonparametric tests, the Mann-Whitney U test is the more powerful of the two

SEM

is not exploratory at all; instead, it is used to assess the researcher's conceptualization, or model, of causal relationships dictated by theoretically-based hypotheses. Moreover, SEM can accommodate the simultaneous testing and comparison of multiple models.

Mediator Variable

is the result of an exogenous variable having an influence that passes through an endogenous variable (the mediator) before affecting another endogenous variable

SEM

is used in an effort to illuminate any causal connections that may exist among a study's unseen variables.2 Once an SEM data analysis is completed, these causal links between variables frequently are displayed pictorially by means of a path diagram.

McNemar's chi-square test

is very much like a correlated-samples t-test in that two sets of data being compared can come either from a single group that is measured twice (e.g., in a pre/post sense) or from matched samples that are measured just once

Multiple R

is what we get if we compute Pearson's r between Y and scores for the individuals who provided scores on the independent and dependent variables.

Point Estimation

no level of confidence must be selected,no statistical table must be consulted,and no interval must be created. Instead, the researcher simply computes the statistic on the basis of the sample data and then posits that the unknown value of the population parameter is the same as the data-based number.

Two Way Repeated Measures ANOVA

requires that each participant travel across all cells created by the two factors,with each person being measured once within each cell

Sphericity Assumption

says that the population variances associated with the levels of the repeated measures factor, in combination with the population correlations between pairs of levels, must represent one of a set of acceptable patterns. One of the acceptable patterns is for all the population variances to be identical and for all the bivariate correlations to be identica

Two Way Repeated Measures ANOVA

second main difference between two-way ANOVAs with and without repeated measures involves the ANOVA summary table

df values (one way ANOVA)

simply counting the number of groups,the number of people within each group, and the total number of participants—with 1 subtracted from each number to obtain the df values presented in the table

pairwise comparisons

simply means that groups are being compared two at a time

Bi-variate correlation magnitude

small relationship r <.35 medium relationship .35 > r < .50 large relationship r > .50

Odds Ratio

sometimes reported as OR, and it is analogous to in that it measures the strength of association between the independent variable and the study's dependent variable

Alpha

specified the probability of making a Type I error

Power

specified the probability of not making a Type II error

Simultaneous Multiple Regression

the data associated with all independent variables are considered at the same time. This kind of multiple regression is analogous to the process used in preparing vegetable soup where all ingredients are thrown into the pot at the same time,stirred,and then cooked together

Logistic Regression

the focus is on the noncontrol independent variables, with the goal being to identify the extent to which each one plays a role in explaining why people have the status they do on the dichotomous dependent variable

Stratified Random Sample (Probable)

the population must first be subdivided into two or more parts based on the knowledge of how each member of the population stands relative to one or more stratifying variables. Then, a sample is drawn that mirrors the population percentages associated with each segment (or stratum) of the population. Thus, if a researcher knows that the population contains 60 percent males and 40 percent females, a random sample stratified on gender should contain six males for every four females.

Model Respecification

the researcher has two goals. One goal is to have a revised model that fits the data better. The other goal is to have a modified model that is parsimonious

Snowball Samples (Non Probable)

the researcher locates a part of the desired sample by turning to a group that is conveniently available or to a set of individuals who possess certain characteristics deemed important by the researcher. Then,those individuals are asked to help complete the sample by going out and recruiting family members, friends, acquaintances, or coworkers who might be interested (and who possess, if a purposive sample is being generated, the needed characteristics

Confirmatory Factor Analysis

the researcher plays a more active role than in an exploratory factor analysis. Guided by theory or the findings from previous research, the researcher in this kind of analysis specifies, on the front end, the desired number of factors and how measured variables are related to those factors. Returning to our museum metaphor, the researcher here is like a person entering a museum that has been previously visited by one of the researcher's friends. That friend has described the museum's floor plan, where the best pieces of art are located,when the crowds will be gone,and how to join a docent-guided tour. Armed with this information,our museum visitor has various expectations as to what things are on display, where they are located, and how to navigate around the museum

Simple Random Sample (Probable)

the researcher, either literally or figuratively, puts the names of all members of the population into a hat, shuffles the hat's contents, and then blindly selects a portion of the names to determine which members of the total group are or are not included in the sample. The key feature of this kind of sample is an equal opportunity for each member of the population to be included in the sample

Moderated Multiple Regression

their goal is to see if the findings of the multiple regression are the same (or perhaps different) for different subgroups of people or different settings

One-way between groups ANOVA

to draw inferences about whether a set of groups all has equal population means. There is only ONE independent variable. IV or factor is categorical. Between groups when you have independent groups.

Logistic regression purpose

to predict scores on a categorical DV from scores on a set of IV's. IV can be noninal, categorical or continous

Observed Variables (SEM)

variables that can be measured. Examples of such variables include your pulse rate, the score you earn on a test, your age, the number of siblings you have, how frequently you blink during a 60-second interval, and how many calories you typically consume in a day. In

Wald Test

was used to see if the odds ratio was statistically significant. This test is highly analogous to the t-test in multiple regression that is used to see if a beta weight is statistically significant

Hierarchical Multiple Regression

we would hold back some of the vegetables (and not put them in with the others) if they are tender to begin with and we want to avoid overcooking them

Post hoc comparisons

were developed because a one-way ANOVA F, if significant, does not provide any specific insight into what caused the null hypothesis to be rejected

Planned comparisons

were developed because researchers sometimes pose questions that cannot be answered by rejecting or failing to reject the null hypothesis of the more general one-way ANOVA .

Practice Effect

with participants performing better as they warm up or learn from what they have already done

Fatigue Effect

with participants performing less well on subsequent tasks simply because they get bored or tired

Confounding

with things that the participants do or learn outside the study's setting between the points at which the study's data are collected