Term
|
Definition
| A minimum score for each selection technique used. If the applicant scores lower than the cutoff on any one of the multiple selection techniques used, then they will be removed from the applicant pool. This is good for any position that requires multiple critical skills and the lack of any of them can be detrimental to performance on the job. |
|
|
Term
|
Definition
| In the multiple hurdle model, applicants must pass the minimum cut-off for each predictor in succession. Essentially, they must complete one task to a level of satisfaction before the next one can be attempted. It's similar to the multiple cutoff model, but fewer applicants will be screened because the amount of applicants at each successive level decreases because they are eliminated if they do not properly complete the assessment. |
|
|
Term
|
Definition
| Multiple regression is designed to analyze the relationship between multiple predictor or independent variables and a criterion variable or a dependent variable. In linear regression, there is only one predictor. Here, there are multiple predictors used in determining the criterion. Each predictor is assigned a different weight due to having a different strength in predicting the criterion variable. Also a compensatory model. |
|
|
Term
|
Definition
| A nonlinear decision model is one that does not use a regression equation to predict the criterion. Multiple cutoff and multiple hurdles are examples of nonlinear decision models. |
|
|
Term
| Pearson correlation coefficient |
|
Definition
| The Pearson correlation is a measure of strength of a linear relationship between variables X and Y. Values are between -1 and 1. 0 would indciate no relationship between the two variables. 1 indicates a perfect positive relationship. -1 indicates a perfect negative relationship between the two variables. |
|
|
Term
|
Definition
| The predictor cutoff typically defines the relative point for scores on some kind of criterion predictor where an employment decision would be made. If an individual scores above the predictor, then they would likely be given employment. If they score below the predictor, then they very well might be rejected in terms of an employment decision. |
|
|
Term
| Pure Clinical Decision Model |
|
Definition
| A pure clinical decision model is a type of job selection procedure that uses both judgmental data collection and judgmental data combination to make a decision. For example, an employer might read a potential employees references and use an unstructured interview. The individual did well in both of those regards so the employer makes the arbitrary decision to accept the employee. |
|
|
Term
|
Definition
| A quota system requires that a fixed number or ratio of people be put into jobs based on their group status. They might not be considered as much on merit as opposed to their group status. Quota systems are used when certain demographics are trying to be achieved. |
|
|
Term
|
Definition
| In a regression model, the intercept is the point where the regression line of best fit crosses the y-axis. |
|
|
Term
|
Definition
| Regression slope is the steepness of a regression line in a quantified fashion. The slope itself equals the change in Y for each unit change in the variable X. One way of expressing slope is Y/X. If slope is positive, Y increases as X increases. If slope is negative, Y decreases as X increases. |
|
|
Term
| Sequential decision strategy |
|
Definition
| A sequential decision strategy is where decisions are made step by step before going on to the next one in the sequence. There could be multiple steps or multiple options in a sequential decision strategy. There is a high risk here for type I error due to the amount of decisions that might be made. |
|
|
Term
|
Definition
| The shrinkage problem is when regression weights from one sample are likely to be smaller when used with a second sample from the same population. Regression weights are somewhat dependant on the sample they are obtained from, thus they might not be applicable to another sample. Shrinkage is likely in situations where initial validation samples are small, a "shotgun" or more scattered approach to collecting the sample is used, or the number of predictors is increased. Shrinke is less likely when the selection items are chosen on the basis of established psychological theories. |
|
|
Term
|
Definition
| A test batter is a set of tests used for a selection assessment. These tests are typically trying to predictor different areas of a performance domain. Test batteries can either be fixed or flexible in nature. |
|
|
Term
|
Definition
| Top down hiring is the hiring of specific employees that are trained and qualified for specific jobs. These employees are very skilled at what they do and are harder attract than a more general hire. This is different from bottom up hiring. |
|
|
Term
|
Definition
| Trait ratings are used when judgmental data is collected, but they are assigned ratings or quantified through some way. Data are combined through predetermined rules that must be used mechanically. |
|
|
Term
|
Definition
| A process in a linear model in which all of the weights are assigned values of 1.0. The is done in a variety of contexts. For example, unit weighting is appropriate when populations change from time to time. It's also useful when a linear composite is used to boost effect size. In general, unit weights are useful in simple or linear regression when the sample size is small. |
|
|
Term
| Univariate vs. multivariate |
|
Definition
| These are the two different types of linear regression. Univariate regression merely uses one predictor to analyze the criterion variable. Multivariate regression uses two or more predictors to more accurately capture the different facets of the criterion variable that might not be covered by a single predictor. |
|
|
Term
|
Definition
| In an experiment or study, the dependent variable is the one being measured. Experimenters are looking at how the manipulation of the IV changes the result of the dependent variable. |
|
|
Term
|
Definition
| Face validity essentiall describes if something is a valid measures of a construct as face level. If a measure of intelligence looks like it measures intelligence, then it is face valid. |
|
|
Term
|
Definition
| Generalizability theory is an expansion of classical test theory. In test theory, there is merely just a true score and an error term in response to an individual's obtained score on a measure. In generalizability theory, there are multiple errores that respnd to different things such as fatigue, environment, test giver, or test form. You might take a single measurement of error, but generalizability theory says you can break this single measure of error into different measures corresponding to different things such as the aforementioned. |
|
|
Term
|
Definition
| Test banding is used to group a bunch of different test scorse that are close together in range to make an easy judgment based on their predictor scorse. Typically, a range of similar predictor scores will be created into a band. This band can be used to make decisions potentially affecting the criterion decision. |
|
|
