Fall 2006 Comp Exam
Terms
undefined, object
copy deck
- Achievement Tests
- These are used to measure knowledge in a specific content area, such as math or reading. They are the most commonly used tests when the outcome being measured is learning and they can also be used to measure the effectiveness of the instruction that accompanied the learning.
- a norm-referenced test
- where an individual’s test performance is compared to the performance of other individuals.
- criterion-referenced test
- where a specific criterion or level of performance is defined, and the only thing of importance is the individual’s performance on the test without any relation to other test takers.
- Standardized tests
- are usually produced by commercial publishers and have broad application across a variety of different settings. They use a set of instructions and scoring procedures that are standard and to be used by all who administer or take the test.
- Researcher-made tests
- are designed for a much more specific purpose and are limited in their application to a much smaller number of people.
- Analysis of Co-Variance (ANCOVA)
- is an inferential statistical tool used to analyze the effects of two (2) or more independent variables simultaneously within the same research design and to determine interactions among variables in multiple sample groups. It equalizes any initial differences between groups and makes groups equivalent with respect to one or more control variables. On its simplest level, it subtracts the influence of the relationship between the two (2) co-variants and the dependent variable from the effect of one treatment. This is similar to handicapping un-equals to make them more equal. Analysis of Co-variance is a procedure for determining whether the difference between the mean scores of two (2) or more groups on a post test is statistically significant after adjusting for initial differences between the groups on a pre-test.
- Analysis of Variance (ANOVA)
- A powerful parametric test of statistical significance that compares the means of two (2) or more sets of scores to determine whether the difference between them is statistically significant.
- Cause and Effect
- A relationship determined only by experimental research in which only the factors which you directly want to measure are isolated by eliminating all other factors that might be responsible for the outcome. When A is done to B it equals C and always does so under the exact same conditions.
- Central Limit Theorem
- The theorem in inferential statistics that states that regardless of the shape of the population distribution repeated samples from it will produce means that are normally distributed. It is the basis or inferential statistics and is the critical link between the results you get from a sample and being able to generalize then to the population. One of the keys to the successful operation of this theorem is that the sample size be greater than 30. If the sample size is less than 30, you may need to apply nonparametric or distribution-free statistics. This theorem is important as it illustrated how powerful inferential statistics can be in allowing decisions to be based on the characteristics of a normal curve when indeed the population from which the sample was drawn is not normal. This fact alone provides enormous flexibility and in many ways is the cornerstone of the experimental method. Without the power to infer, the entire population would have to be tested. an unreasonable and impractical task
- Chi-Square
-
A nonparametric inferential statistical test technique used to compare two or more groups on a nominal variable with two or more categories. Observed frequencies are compared with expected frequencies. This form of testing has been recommended as the most appropriate statistical procedure for analyzing categorical variable that are exclusive, independent, and exhaustive. That is, either nominal (refers to grouping on the basis of observed, qualitative distinctions such as gender, racial/ethnic background, or type of school program) or ordinal (refers to rank-ordered data where the assignment of numbers reflects degree rather than kind, such as incremental categories of income, educational aspirations, or age group).
Chi-Square tests provide either a measure of (1) the goodness of fit for the distribution of a single variable when compared to some theoretical distribution, or (2) the association or relationship between two variables. It yields an inferential statistic known as Chi represented by the symbol X2. It is the only test that can be used when both independent and dependent variables are qualitative in nature. - Coding.
-
This is a technique that allows data to be put in a format that lends itself to data analysis by assigning numbers to represent data. Example: male =1 female =2
When coding, the simpler the values, the less confusing and the better the data analysis becomes. One rule for coding data is to use codes that are reduced in clutter and as unambiguous in meaning as possible without losing the true meaning of the data themselves. Make them as explicit and as discrete as possible and do not combine data during the analysis process. - Continuous Recording
- Is simply recording behavior on a continuous basis. It is where all the behavior of the target subject is recorded with little concern as to the specificity of its content. The behaviors recorded are those that occur in the natural stream of events. It is a broad and fruitful way of collecting information, but since little planning goes into what is recorded the information requires intensive sifting and sorting at analysis time.
- Continuous Variable
- A variable that for analysis purposes can be fitted on a numeric continuum. Each participant can be assigned a numeric score that is a point on the continuum (age, achievement score, monthly income, etc). Included among this categorization of variables are measures on interval, ratio or Likert type scales.
- Control Variable
- This is a variable that has a potential influence on the dependent variable. This influence has to be removed or controlled to neutralize its effect on the dependent variable. Control by manipulation is possible only in experimental studies. Example: When studying reading speed and comprehension, intelligence will also have an effect, so one must control for intelligence and make it a constant (neutralize its effect).
- Correlation.
- This is the relationship between two or more variables. It is how two or more things are related or how a specific outcome might be predicted by one or more pieces of information. It does not imply a cause and effect. The higher the correlation the higher the degree of relatedness. In a direct correlation the variables change in the same direction with each other. In an indirect correlation, as one variable goes up the other goes down or visa versa. It is the absolute value of the correlation coefficient, not its sign that is important when determining how highly the variables are correlated.