Presentation on theme: "Ex Post Facto Experiment Design Ahmad Alnafoosi CSC 426 Week 6."— Presentation transcript:
Ex Post Facto Experiment Design Ahmad Alnafoosi CSC 426 Week 6
Ex Post Facto what??? Webster Dictionary defines Ex Post Facto as: after the fact : retroactively Late Latin, literally, from a thing done afterward. First Known Use: 1621
Explain More… In situations where it is not possible to manipulate variables. Ex Post Facto design provides an alternative to investigate how independent variables affect dependant variables. The researcher can observe the independent variables after the event.
That sounds like Co-relational design? Co-relational design and Ex post facto design involve examining existing conditions. Ex Post Facto design has dependant and independent variables whereas Co-relational design does not.
What about experimental Design? Both experimental design and Ex post facto design have independent and dependant variables. Ex Post Facto differs that it does not introduce the presumed producing cause. Thus in Ex Post Facto the researcher is NOT able to draw firm cause and effect. Both share similar designs.
What does Ex Post Facto Design Look like Similar to Experimental design, ex post facto design has multiple forms. These form involve variation of events (experience), Observations, Groups and combination of the above.
Simple Ex Post Facto Design
Similar to Static Group Comparison with the difference of the timing of the treatment (Experience). It is called Experience since the researcher can not control it. Association can be drawn from this study (NOT Cause and effect).
Factorial Design In designs that involve multiple dependant variables with Ex Post Facto design, Factorial design is needed.
Randomized Two Factor Design 2 variables tested by 4 groups. Variable 1 effect can be studied by comparing group1 and group2 of that of group3 and group4. Variable 2 effect can be studied by comparing group 1 and group 3 of that of group 2 and group4
Randomized Two Factor Design - Cont This design is a generalized version of Solomon four group design. (event instead of experiment) This design allow to see the effect of each of the variables. It also can show the interaction effect of the variables.
Combined Experimental and Ex Post Facto Design Combining experiment with Ex Post Facto Experience It has Ex Post Facto component by initially selecting groups that have that experience. Then there is experimental phase where where experiment is conducted.
Combined Experimental and Ex Post Facto Design - Cont The results will be 4 groups all possible combinations of experience and experiment. This design enables the study of experiment effect the dependant variables Also it enables the study of how previous experience interact with the experiment.
Sampling Ahmad Alnafoosi CSC 426 Week 6
Choosing a Sample in Descriptive Study The purpose of descriptive study is to be able to determine and describe large population. In most instances surveying all the population is not possible because of the sheer size. On the other hand the sample needs to be large enough to be representative of the population and their characterizations that are relevant to the study.
Sampling Design To achieve the aforementioned goals a sampling design is needed. The sampling design needs to take into consideration the actual traits of the population to apply the appropriate sampling design.
Probability Sampling Researcher can specify that each segment of the population will be represented in the sample. The sample is chosen using Random Selection (each member of the population has equal chance to be picked)
Simple Random Sampling Is a probability sampling design. Each member has equal chance to be picked. Used for small population where every member is know.
Stratified Random Sampling Is a probability Sampling design. Is used in stratified population where there is multiple layers strata Guarantee that each of the identified strata. Is used when the stratum are equal in size.
Proportional Stratified Sampling is probability sampling design. When the population is stratified but where stratum are not equal in size. In this case the number of random sample of each strata taken is dependant proportionally to the strata population to the whole population.
Cluster Sample is probability sampling design. Is used when the population is spread over large area. Clusters need to be similar to each other as much as possible. Each cluster has to have equal heterogeneous population.
Systematic Sampling Is probability sampling design. Involve selecting individuals based on pre- determined sequence. The sequence needs to be random.
Factors in determining Probability Sample Design Population size Stratification Size of stratum Clustering
Non-Probability Sampling Does not guarantee that each element of the population will be represented in the sample. Some members of the population have no chance of being represented.
Convenience Sampling Is non probability sampling design. AKA Accidental sampling. It sample available members of the population.
Quota Sampling Is Non-probability Sampling It select individuals in the same proportion as they are found in the general population, but it is not random.
Purposive Sampling Is Non-probability Sampling It select individuals for a particular purpose. Needs to be careful since it assume that the chosen sample is useful for the purpose.
Sampling Surveys of very large population To tackle very large population multi- staging of sampling areas might be needed. This involves Primary area selection Sample location selection Chunk selection Segment selection Housing selection
What is the right sample size For population less than 100, sample the entire population. For population around 500, sample 50% For population around 1,500, sample 20% For population larger than 5,000 sample size can be around 400.
Sample Bias Sampling will introduce bias into the sample. Researcher need to acknowledge.