Social work practice and professional applications

Social Research

There are several explanations for an observed relationship that need to be examined and dismissed before a true relationship can be established. First, it is possible that an observed relationship is merely coincidental; using large sample sizes that seem to show the same relationship can diminish this risk. Also, an external cause influencing both of the observed factors could exist, which is why testing or controlling all other variables is necessary to show the validity of a relationship. Finally, it is possible that a researcher bias exists, and again controlling the research methods and variables can help to limit exposure in this area.

A Type I error occurs when the null hypothesis is falsely rejected. In this type of error, the correlation, relationship, or explanation advanced by the alternative hypothesis is accepted -- meaning the null hypothesis is rejected -- due to false observations and/or analyses. A Type II error occurs when the null hypothesis is not rejected even though evidence that it should be rejected actually exists. For instance, the correlation proposed by an alternative hypothesis might actually exist, but due to an error this correlation is not observed or is not seen as significant by the researcher, resulting in continued acceptance of the null hypothesis.

3)

A null hypothesis would be that there is no correlation between caseload size and child welfare worker stress. A one-directional research hypothesis would be that as the size of a caseload increases, the amount of stress experienced by the child welfare worker with this case load also increases. In the research that would investigate this hypothesis, caseload size would be the independent variable or the cause in the causal relationship hypothesized, while child welfare worker stress would be the dependent variable, or the effect.

4)

Probability sampling is also known as random sampling, and can be used to infer information and observations about the general population from which the sample is drawn. In other words, a large enough random sample population drawn from a general population pool would be expected to match that general population in demographics and proportion, and thus the probability of a specific outcome in the population would be the same or similar to observations in the general population. Nonprobability sampling can be useful, but is not entirely random and so cannot be used to make assumptions about a general population.

5)

Some advantages of questionnaires are the ability they provide for respondents to give detailed answers, and the removal of the researcher from the direct data gathering process (i.e. respondents can fill out questionnaires in their own time) reducing research bias or influence on the results. Disadvantages, however, include the potential for misinterpretation or differing interpretations among respondents, reducing the validity of the questionnaire's measures and conclusions. Reduced control generally can be a problem.

6)

Interviews have a strong advantage in the ability of the researcher to clarify questions and gain deeper insight into responses, ensuring accurate and meaningful interpretation on the parts of both the researcher and the respondents. A disadvantage to interview research methods, on the other hand, include the potential for bias based on the respondents' desire to please the researcher, who will necessarily be directly involved in the research process.

7)

In the One-Group Pretest-Posttest experimental design, a single group of subjects is given a test that measures a certain variable or variables, then is exposed to the experimental condition, and then is tested again to measure the effects of this exposure. While this might seem valid, there are several threats to internal validity facing this design. Other intervening events could affect posttest scores, the simple passing of time could have an effect, different test conditions could lead to different results, and selection biases could also influence outcomes.

8)

There are a number of factors that influence the sample size needed for a given piece of research. The desired confidence level of the research outcomes, the level of precision necessary in the data collection and analysis, and the degree of variability that exists in the population at large can all have an effect on the needed sample size. For greater levels of precision and confidence, larger sample sizes will be needed. Greater levels of variability in the general population would also require a larger sample size, in order to ensure that the population is well and accurately represented across all of its variability.

9)

Random sampling or selection refers to the selection of subjects from a general population pool. This step is taken at the very beginning of research to ensure that the sample population is representative of the general population. Random assignment refers to randomly placing selected subjects into various research groups. This step is taken in order to ensure that no bias is created in different experimental and/or control groups, so that each group is comprised of similar proportions and demographics.