Term
|
Definition
| include the important predictors in your model without attempting to add every explanation |
|
|
Term
|
Definition
| the simplest explanation is usually the best one |
|
|
Term
|
Definition
| those in which the coding of the variable is a close reflection of their meaning |
|
|
Term
|
Definition
| (interval) can assume any real value, i.e. temperature |
|
|
Term
|
Definition
| (count data) can only assume integer values, cannot be divided into fractions, i.e. number of children |
|
|
Term
|
Definition
| those whose numerical representation in the data set simply classifies mutually exclusive values of the variable, but those numerical representations are substantially less meaningful than what is found with quantitative variables |
|
|
Term
|
Definition
| qualitative variables in which the numbers represent a hierarchy of values. i.e., attitudes or grade in school |
|
|
Term
|
Definition
| qualitative variables in which the numeric values assigned to values are merely placeholders, i.e. race |
|
|
Term
|
Definition
| (binary variables) variables in which there are two and only two possible outcomes |
|
|
Term
|
Definition
| binary variables that have been coded with 0 and 1 |
|
|
Term
|
Definition
| new variable created by adding up a number of variables together |
|
|
Term
|
Definition
| new variable created by combining (sometimes weighting) variables to create something new altogether, i.e. socioeconomic status |
|
|
Term
|
Definition
| underlying construct of the components in a scale |
|
|
Term
|
Definition
| indirectly measured construct, i.e. socioeconomic status |
|
|
Term
|
Definition
| a condition created when two or more variables measure the same underlying construct |
|
|
Term
|
Definition
| refers to any one of a number of mathematical techniques for transforming variables so that they are in a single metric |
|
|
Term
|
Definition
| 2 or more variables are measuring the same thing (score on one variable should predict the score on another variable0 |
|
|
Term
|
Definition
| measures the internal consistency of several variables. Alpha reliability ranges from 0 to 1, where 1 is perfect internal reliability and 0 is the variables do not measure the same thing at all. (alphas greater than .7 are generally considered to have enough internal relaibility that scaling is warranted |
|
|
Term
| exploratory factor analysis (purpose) |
|
Definition
1) can be used to see if a single variable can be used to replace multiple variables
2) can determine if several variables can be scaled together
3) detect clusters of variables |
|
|
Term
| 3 types of factor analysis |
|
Definition
1) Exploratory factor analysis (EFA)
2) Principal component analysis (PCA)
3) Confirmatory factor analysis (CFA) |
|
|
Term
| Pearson's correlation coefficient |
|
Definition
| used to understand how variables are related to each other; measured from -1 to 1 (negative number=negative association; positive number=positive association) |
|
|
Term
|
Definition
| produced by EFA; measures the combination of the linear composites and the structure coefficients |
|
|
Term
|
Definition
| in a factor matrix, the amount of weight you would give a variable if you were going to put it in a scale |
|
|
Term
| 3 stages of factor analysis |
|
Definition
1) Selection
2) Extraction
3) Rotation |
|
|
Term
| Selection stage of factor analysis |
|
Definition
| determine the number of factors that are meaningfully represented by the variables |
|
|
Term
|
Definition
| in factor analysis, indicates the variance accounted for by each underlying factor |
|
|
Term
|
Definition
| in selection of FA, the total number of factors represented by a set of variables is equal to the number of eigenvalues greater than 1 |
|
|
Term
|
Definition
| in selection stage of FA, a graph with the factors on the x axis and the eigenvalues on the y-axis. The scree that occurs after the curve bends is simply error variation and does not represent true factors |
|
|
Term
|
Definition
| 2nd stage of FA, the process by which we account for variance |
|
|
Term
| best method for FA extraction |
|
Definition
| principal axis factoring with iterated communalities (least squares)- best approach, especially if smaller sample size |
|
|
Term
|
Definition
| 3rd stage of FA, extremely complex mathematical procedure that requires you to think in crazy dimensions. Basically, it rotates the axis to place it nearer to the data points |
|
|
Term
|
Definition
| Type of rotation in FA; assumes factors are uncorrelated to each other |
|
|
Term
|
Definition
| type of rotation in FA; assumes correaltion |
|
|
Term
| Rule of thumb when determining which variables load on which factors in FA |
|
Definition
1) factor should have at least 4 variables
2) each variable should load close to or higher than +/- .5 |
|
|
Term
|
Definition
| a measure of whether the data are peaked or flat relative to normal distribution |
|
|
Term
|
Definition
| mean and standard deviation |
|
|
Term
|
Definition
| show bivariate differences of means or proportions |
|
|
Term
which inferential statistics to use if 1)using percents or proportions
2)means |
|
Definition
1) Pearson's Chi-Square
2)Student's t-test or an f-test derived from ANOVA |
|
|
Term
|
Definition
| how most nominal and ordinal data are entered into models; a dummy variable is created for each category of the variable |
|
|
Term
|
Definition
used to compare 2 (and ONLY 2) means to see if they are statistically different from each other.
Most often use the two-tailed t-test where Ho= means are equivalent and Ha= means are different. First, must test for equal variance. p< .05 means we are 95% sure that the difference in means observed is not simply due to sampling error. |
|
|
Term
| Robust test for equal variance |
|
Definition
| used before a t-test can be conducted to determine whether the two groups have equal variance on the variable under question. Want p > .05 |
|
|
Term
| Levene's test for equal variance |
|
Definition
| most common used robust test for equality of variance; (w0) ONLY indicates whether you have equal variance, not if you have statistically significant differences of means. Ho= equal variance; ha= unequal varaince. want p> .05 |
|
|
Term
|
Definition
| (Analysis of variance f-test): used when comparing the means of three or more groups. One-way ANOVA- ho= all means are equal; Ha= at least two groups have different means. If p is significant, then must conduct a post-hoc test |
|
|
Term
| 3 types of post-hoc tests |
|
Definition
Used after geting a significant p value in a one-way ANOVA.
1) Bonferroni
2) Scheffe
3) Sidak |
|
|
Term
| Barlett's test for equal variance |
|
Definition
| Addresses the major limiting assumption of ANOVA which assumes that the population variances are equal. Ho= equal varainces, Ha= unequal variances. Pant p > .05 so we fail to reject the Ho |
|
|
Term
|
Definition
| a statistical technique used to measure and describe a relationship between two continuous variables. |
|
|
