An independent samples t-test is very much like a comparison of two confidence intervals. ( A confidence interval being used to indicate the reliability of an estimate.) If the CI overlap, then there is not enough evidence to declair that the groups are different. If the CI do not overlap, then the two groups are very likely to be different.

In a boxplot, the distributions may look very similar. An independent samples t-test would suggest whether the difference between the means is significant or not (p<.05). In an independent samples t-test, a p-value greater than .05 indicates that the means are not significantly different. A p-value less than .05 indicates that the difference between the two groups is significant.

Continuous data: to compare the means of two or more gorups on a continuous outcome (at one point in time); to estimate the effect of a treatment on an experimental group compared to a control group; to determine the extent to which variation in an outcome can be attributed to membership in a particular group.

ANOVA is the partitioning of variation in an outcome into within- and between-groups components in order to determine the statistical significance of differences in group means.

One-way ANOVA allows us to test for differences between groups on one factor.

An F-test will determine whether the difference in means is significant, but it will not tell you which groups. The Tukey's Post Hoc test is frequently used to compare the means in various groups to determine which groups varying significantly.

Each observation must be independent of the other observations in the data set.

The relationship between two variables ranges from a perfect inverse relatiionship (r=-1.0) to no relationship (r=0) to perfect direct relationship (r-1.0). The rule of thumb for interpreting correlaton is weak < .20; moderate .20-.50; strong > .50.

Correlation does not prove causation because of (1) directionality (Does A lead to B or B lead to A?); (2) third variable (C caused both A and B and their correlation.); (3) chance (What if the observed relationship occured only by chance?)

To test the significance of the model, an f statistic is necessary. This is used to determine a p-value (the significance of whether the independent variables are significant predictors of the outcome variable.

In multiple regression analysis, the coefficient tells you how much each predictor variable is expected to increase when the predictor variable increases by one while holding the other predictor variables constant. When you report your interpretations of the results, you need to include the unit of measure because each predictor variable will likely use different units of measure. Basically you need to do this for clarity, accuracy and specificity. It is important information in the interpreting and reporting process.

Linear model: Using data from the standardized reading test, how much does achievement go up as SES increases?

Multiple model: What is the impact of SES, ethnicity, and participation in the arts on grade point average.

ANCOVA test for differences in post-test performance after controlling for or holding constant pre-test performance.

ANCOVA can include more than one continuous variable as a covariate in the model. Imagine a model where you wanted to compare post-test performance after controlling for students' pre-test performance and SES.

MANOVA is simply an extension of standard ANOVA where the assumption that each observation is independent is relaxed. In other words, the model adjusts for the fact that we are making multiple measurements on the same person, and it estimates the correlation between these measurements.

MANOVA can also be extended to include continuous covariates as control variable. This is called MANCOVA.

Present the data for the variables in a contingency table.

Compare the actual counts in each cell to the counts that would be expected if each variable was independent. Do this by multiplying the row and column totals by the overall proportion of individuals in each group.

Compare the observed values to the expected values to see if there is significance to the difference between observed and expected.

If there is significance, then there is evidence to suggest that being in one category on one variable is associated with being in one category on another variable.

If there is no significance, then the observed difference is likely by chance.

Constructs are intangible and typically measured indirectly through a collection of relevant survey items.

For survey scales, rules of thumb: 1.00 is perfect; betweeon .90 and 1.00 is very good; between .80 and .89 is good; between .70 and .79 is moderate; between .60 and .69 is low; below .60 is very low.