Trend studies: use sequential cross-sectional A large group of volunteers agreed to
surveys to determine how population report details about their lifestyle,
characteristics are changing. including such things as eating and Is the rate of crime increasing in the exercise habits. United States? This info is tracked throughout their Cohort Studies: use the same cohort of lives and used to develop behavioral individuals over time. profiles, which have been found to have A cohort is a group of people who falls an effect on the health of a person’s within a defined time range. heart. A cohort study can be developed where Once begun, a panel study can only a cohort is asked the same series of include those individuals included in the questions periodically. original panel. Be sure that the members of the cohort To continue the research beyond the life studied subsequently were people who of the original panel, as is being done fit the original cohort definition. with the Framingham Study, researchers Ex) The change in the level of happiness of must restart the research periodically married couples with new panels of participants We may establish cohorts of those Cross-Sectional Surveys married in the 2000s, such as those Collect data from a cross-section, or a between the ages of 26 to 35. representative segment, of a population We want to see how their attitudes at a point in time changed over time. Take a snap shot of the topic Survey them every ten years. During the They are the types of surveys most second survey, this group would be 36 to frequently used. 45 years old. Criminal Justice Research Panel Studies: are used to identify changes Consumer Behavior Research in the same set of people on two or more Voting Preference occasions. Political Surveys: With only two weeks Ex) Framingham Heart Study: is now to go, the race for Sheriff is a statistical recruiting its third generation of participants dead heat. The complexity of the information 7 Guidelines for Creating Effective Questions sought; Complex information should be and Questionnaires collected through interviews rather than 1. Conciseness; Questions should be as with a questionnaire. short as possible. Saving a few words on The sensitivity of the information every question can make the sought; Sensitive topics, such as mental questionnaire easier to read and and physical health matters, sexuality complete issues, or deviant behavior are best dealt 2. Clarity, Questions should be clear. with by techniques that ensure the Questions can almost always be edited to respondent will be kept anonymous make them clearer. This will make Anonymity is when the identity of a questions easier to comprehend and respondent is not known by anyone ensure that respondents interpret the Confidentiality is when the identity questions the same way of a respondent is known by the 3. Jargon and Slang, avoid using jargon or researcher, but will not be made slang. The level of wording may known to anyone else under any alienate, offend or frustrate respondents. circumstances Jargon is the use of words that are The literacy level of the study frequently familiar to a research and population; If you are unsure of the those in the same field of study. Slang is literacy levels of your population, you the use of words whose meanings may should also consider using interviews be offensive to respondents rather than questionnaires 4. Double-barreled Questions, all questions The time and cost of data collection; should inquire about one topic at a time. Large groups can often complete a A double-barreled question is two lengthy questionnaire in 15 to 20 questions in one minutes. The cost of distributing 5. Sequencing Questions, simple and less questionnaires threatening questions should precede The geographical area of the study complex and more threatening questions population; The least expensive way to 6. Leading Questions, question wording contact a dispersed study population is should not bias or lead the respondent’s by mailed questionnaires answer 7. Sensitive Questions should be asked in a Single 3 or more Single relatively non-threatening way sample sample sample mean. Response Rate proportion. means. Face-to-face interviews have fairly high Two sample Two sample response rates => About 80 percent or proportions. means. higher Mass-administered questionnaire Before and surveys also have quite high response after, or rate => sometimes as high as 95% paired. Mailed questionnaire surveys usually have the lowest overall response rate => We use traditional experimental approach as about 30% a device for comparison. To boost the rate of response, Appropriate for hypothesis testing researchers frequently try to provide Strengths—Inference incentives such as enclosing dollar bills Weakness—generalizability or lottery tickets. Three criteria for the classical Choosing a Measure of Central Tendency experiment Level of measurement IV and DV Interval-ratio: mean, median, mode Pre-Post testing Ordinal: median, mode Experiment and control groups Nominal: mode Experimental and Control Groups Shape of the distribution, if you have Control group = non-treatment group interval-ratio data, but the distribution is Experiment group = treatment group very skewed, then the median is usually a better choice than the mean.
Z-score t-test ANOVA
Double-Blind Experiments Construct validity is concerned with Control group = taking placebo generalization from our observations to Experimental group = taking a new drug actual causal processes in the real world, Both researchers and subjects are so the concept of construct validity is “blind” closely related to issues in measurement. Third researcher knows which subjects Since almost any measure of a are in which group. Utilizing this construct is incomplete, the threat to method eliminates the possibility of construct validity always exists in knowing which is the experimental the research. group and which is the control. There are three elements of Threats to the validity: Internal Validity, enhancing construct validity. Construct Validity, External Validity, and The researcher needs to link Statistical Conclusion Validity constructs and measures to The problem of internal validity can lead theory. to inaccurate conclusions about cause The researcher needs to clearly and effect that may be biased due to indicate what constructs are nonrandom or some systematic error. represented by specific measure. History: Historical events may occur The researcher needs to think during the course of the experiment carefully about what levels of that confound the experimental treatment may be necessary to results. produce some level of change in Maturation: Participants are the dependent measure. continually growing and changing. External validity also links to Testing: Participants sense the generalization by checking whether expected outcomes via the process of findings from experiments can apply to testing any person, any time and any place. Instrumentation: Changing testing It is very difficult to ensure external format can cause a problem of validity because we cannot be certain reliability. the result of our findings will have Statistical Regression: outliers in a similar findings in a different group. group can cause inaccurate outcome. Threats to statistical conclusion Four steps for evaluation research: validity can be an issue when findings Problem Identification, development of a are based on small sample size. Since Purpose and Needs Statement to experiments are often costly and time identifies any and all existing and future consuming, many researchers use conditions which inhibit or block the relatively small numbers of participants. attainment of stated Goals and Objectives When randomization is not possible, the Program Planning-Impact Assessment, option would be a quasi-experiment. Impact statements and reports must => the internal validity threats become a present an Alternatives Analysis which significant issue provides a detailed review of how The Non-Equivalent Groups Design various development scenarios might be (hereafter NEGD) is probably the most expected to positively and negatively frequently used design in social research. affect a range of issues The term comparison group is Program Implementation- Program commonly used, rather than control Monitoring, once a program has been group, to highlight the non-equivalence implemented, it must be continually of groups in quasi-experimental designs monitored to ensure it is operating as designed, and to ensure goals and Evaluation research is frequently used to objectives are being met and policies are develop new programs and review the being carried out. effectiveness of existing programs. Program Evaluation Clients who typically require this type of work include: Federal, state, local government Private sector, occasionally when involved with government Government agencies attempting to solve social problems Research Ethics Gunshot Detection System Ethics are guidelines for moral conduct. The use of gunshot detection systems by Dilemmas are conflicts that exist police departments increases police between ideals and societal values. effectiveness in handling gunshot calls Examples of scientific ideals and Shotspotter allows for a better societal values are evident in sensitive description of gun violence issues subjects that are difficult to research: Shotspotter effectively decreases police Corruption in people’s lives response times and times to dispatch. Sexual attitudes/responses Does not statistically improve case Violence outcomes (Action) at any level. One way in which they can improve There are 5 sensitive areas which must be on this is through the combination of considered: CCTV and gunshot detection Potential harm systems. Lack of informed consent, Researchers At the time of this study, Brockton was are required to receive the consent of also looking into placing individual their participants after fully informing Shotspotter “dispatch” screens in police them of the nature of the research and cruisers. how the individual will be participating Possible to further reduce response Deception, the use of deception is a times and times to dispatch common practice. It is difficult to rule Limitation out in certain cases, but is to be avoided Dispatch Log does not contain a running if there is any chance of causing harm to history of each case the participants Although unfounded (Action Type 1) Debriefing, Interviewing participants represents a less desired outcome in this after their involvement in research is study, it does not entirely mean the appropriate and ethical, and may further police were ineffective in responding to benefit the research and the participant the call. Invasion of privacy, Privacy rights must be respected as a basic barrier in our society. Level of measurement Ensuring the validity of indicators. Nominal data can be classified or put Face Validity into categories, but cannot be ranked. This is a subjective assessment the Ordinal is a basic ranking from high to researcher makes in with the results appear low. valid. Due to its inherent subjectivity, it Interval data is numerical data without may or may not be valid. absolute 0 as a starting point. Convergent Validity or Criterion-related Ratio data has an absolute 0 starting validity point. validity is achieved when different approaches to the determination yield the Ensuring the reliability and validity of same answer. (Utilizing State Data + indicators Individual Research Data) Test-Retest Reliability Construct Validity involves administering the same test to the seeks agreement between a theoretical same group twice. concept and specific measuring device or Alternate Form Reliability Technique procedure. Respondents are asked similar questions or # of going out at night => Fear of Crime alternate forms of the question with the Content Validity expectation of receiving the same answer. refers to the degree to which a measure Split-half Reliability covers the range of meanings included With this method, the scale or index is within the concept. divided in half and administered after How safe do you feel walking along in this being fully administered. The results from area after dark? the two are compared. If there is a high Concurrent Validity correlation between two, the measure is This is when a researcher compares the considered reliable. test results with another known indicator. Interrater Reliability High score on the vocational test => is the extent which two or more Good Mechanic individuals agree (coders vs. researcher). This method emphasizes consistency of the implementation of rating system Sampling sampled. The proportion of samples for the process of selecting units (e.g., people, each group = the proportion each group organizations) from a population of interest of the total population. so that by studying the sample we may fairly how middle-school children feel about generalize our results back to the population their school from which they were chosen o Purposive Sampling (or Theoretical Nonprobability Sampling Sampling) refers to the group of sampling techniques a researcher purposely picks extreme in which the probability of being selected cases to illustrate contrasting types or or sampled is unknown. More convenient features to make a point. to use, less expensive, and easier to what makes the best inner-city schools? collect. The accuracy of results is not o Snowball Sampling known, so the reliability and therefore, the starts from a single participant or small validity of the research can’t be asserted. number of participants by asking them o Judgmental sampling to identify others like him or her who a researcher makes an informed guess might be willing to participate in a about a representative sample. study the media on election night qualitative research for gang o Convenience Sampling membership a researcher takes samples from an accessible, unrepresentative large Probability Sampling group. May be useful for pre-testing provides researchers with a variety of questionnaires prior to administering sampling techniques which can be used to them, but the results are seldom develop generalizations. With probability representative of the entire population. sampling, the potential for any member of Using a large class of students or the population to be sampled is equal to conducting interviews at a shopping that of all others. center. o Simple Random Sampling o Quota sampling: all members of the target population a researcher will begin by setting up a have an equal probability of being matrix of important groups to be selected. All of population members are numbered from 1 to N. Random are small pieces of data analysis that capture numbers are used to select those who emergent ideas. will be sampled. Triangulation o Systematic sampling We use more than one method to find the the researcher selects every Kth case same answers or research the same after selecting the first case randomly. conclusions o Stratified Sampling: it may be useful to divide the The five crucial components in field population into strata or subgroups, and research then sample within each group. The Population and Sampling Design purpose of using the function of Data Analysis and Role of Memos stratification is to organize the Data Quality: Validity and Reliability population into homogeneous subsets Access o Proportionate Stratified Sampling Human Subject Protection Issues is a refinement of stratified sampling, that the researcher decides to keep the IRBs (Institutional Review Boards) sample sizes in proportion to the size of Board members make judgments about the each strata or subgroup? overall risks to human subjects and o Cluster Sampling whether these risks are acceptable, given we use multistage samples, where the the expected benefits from actually doing population is divided into primary and the research. subsets. we want to determine They determine whether the procedures to residents’ opinions about their school be used include adequate safeguards system in a large city. Cluster sampling regarding the safety, confidentiality, and can be used to achieve a representative general welfare of human subjects. sample without knowing names or Sample Population residential addresses. If population is homogeneous => 6 to 8 Memo cases When an insight does occur during coding or If population is heterogeneous => 12 to 20 examining field notes or day’s observation, it cases. is essential to get it down in writing. Memos