 |
Glossary A,B,C,D,E,F,G,H,I,J,K,L,M,N,O,P,Q,R,S,T,U,V,W,X,Y,Z Abstract - A shortened summary of your research which is placed at the beginning of your report. This introductory paragraph sums up you research and conclusions in a few sentences. Anonymity - The concept that subjects should not be identifiable as part of a research project. This means that all data should be kept separate from the names of the participants, if those names are kept at all. Association - A link between two phenomena which involves co-occurrence, meaning that when one happens the other probably will as well. It is usually shown through statistics. Association is one of the requirements for the support of a causal hypothesis. Attitudes - What people think is desirable
(as opposed to true or false). Balanced response categories - Responses to questions which have an equal number of response choices on each side of "neutral." Non-balanced response categories are a sign of a biased study. Beliefs - What people think is true and false. Bias - The problem of getting un-representative results because of errors in methodology. Different types include: researcher bias, which can effect survey question design; research question problems; coding problems; and many other problems due to pre-existing thoughts on the topic. See also missing data bias, non-response bias, and prestige bias. Binary choice - A type of question which has two responses to chose between. Bivariate analysis - Case - The subjects that you are studying. Each subject is a separate case, and it's data are put into a row on the statistical software. Cause (causation) - A way to link variables which specifies that a change in the independent variable causes a change in the dependent variable. To prove causation you must have temporal order, association, and the elimination of alternative reasons. Census - Data from the entire population rather than from a smaller sample who you study. Central tendency - Statistics which describe the average of the data. Checklist format - Questions that allow respondents to check all the pre-formatted responses that apply to them. Usually coded like binary choice questions as "checked" or "did not check," but can be coded other ways. Closed question - A question that has pre-formatted answers for respondents to choose from or rank. Cluster sampling - A technique where you randomly select clusters from your population and then take a sample from those clusters. Cluster sampling allows you to more efficiently conduct personal interviews with a dispersed population. Codebook - The entire questionnaire with full questions, the coding scheme, missing data codes, and any other relevant information needed to code the responses. Coding - The process of turning data from the responses (or artifacts) into numerical format, usually for a statistical software package on a computer. Concept - Broad, abstract ideas which are generally defined. Concepts need to be turned into operational definitions before research can precede. Conceptualization - A preliminary step in the research process, conceptualization takes vague, abstract ideas and makes clear, concise definitions so that you may precede to develop your research question. Concurrent validity - A type of measurement validity which uses already formed definitions which are accepted as indicators of the concept. Conditional table - Confidence interval - The range which, based on your sample size and population size, you can say that you have a high confidence that the general population will fall within. Assumes you have a representative sample. Ex. 39% plus or minus 5%. Confidence level - The degree to which you can say that your results are accurate. Based on your population size and sample size. Most scholarly work requires 95+% confidence. Confidentiality - The assurance that all data collected will be held only for the research purposes and then destroyed. Only the researcher will know the identities of the respondents and their specific responses. Data will be presented in an anonymous form to the public. Constant - A variable which does not change in your analysis. Construct validity - One type of validity which can be attained if indicators of a concept match the ideal concept which you are trying to study. Content analysis - A form of non-reactive study which takes artifacts of some type and examines the content to try to learn something about the society and individual which produced it. Content analysis takes the data from the artifacts and analyzes it with quantitative methods. Content validity - Asks whether an indicator of a concept (question, code, etc.) fully represents the concept that it is trying to measure. Contingency question - A type of question that filters people who the question would not apply to. A first question is asked and the respondent should be given clear instructions as to whether they should respond to the next question(s) or skip down to the next main question. Control variable - A variable which is kept the same to examine differences based on other variables. Often used to test relationships between two other variables. Convergent validity - When multiple, similar indicators of a concept are used, convergent validity asks if they get similar results. Convenience Sample - A non-random sample of people who you can find. There is no attempt to be random or representative. The number of people is really the only objective. The problems of non-probablilty sampling come into play here. Correlation - One criteria for proof of causation, correlation is the statistical proof that the two variables exist in relationship with each other. Criterion validity - Uses other criteria (other studies, etc.) to determine whether your measures are valid and indicate the concept correctly. Criterion validity can either be done in a concurrent way, which looks at previous studies and measures to compare yours to, or in a predictive way, which tries to predict the outcome of future studies or tests. Cross-sectional research - Research which looks at multiple groups of people at the same time for analysis (ex. comparing opinions of people in Los Angles to people in Elko, Nevada) Cross-tabulation - Curvilinear relationship - Data - Any attribute of something which you record for analysis in a systematic way. Debriefing - Fully informing your subjects your study. You should reveal the true reason for your study (if it was concealed) and give contact information for yourself and information about getting the subjects a copy of the final report if they wish. Dependent variable - The variable that changes as a result of a change in the independent variable. Descriptive statistic - A statistic which is used to give basic information about your sample rather than infer things about the population. Dichotomous question/variable - A question or variable that has two specific categories (rare because there are usually "don't know" or "neutral" responses possible. Discriminant validity - When multiple, different indicators of a concept are used, discriminant validity asks whether those indicators got different values. Double-barreled question - A bad type of question which asks two things at once. (ex. Do you like baseball and hockey?) Dummy variable Ecological fallacy - A logic error when trying to prove causation, the ecological fallacy occurs when you mismatch levels of data and try to apply statistics at one level to infer to units of another level. Empirical - A necessary quality of a good hypothesis, empirical means that it can be scientifically tested because it is easily observable. Equal weight scales - Good questions should be scaled equally, meaning that the codes on both sides of "neutral" have the same values. Exclusiveness - Exclusive responses to questions allow the respondent to select only one response. If one response is selected, all others are eliminated. Exhaustiveness - Exhaustive responses to questions cover the entire spectrum of possible answers. Explanation - A hypothesis or research question that attempts to explain a concept by relating two variables. External validity - The ability to expand your findings to the general population. Face validity - The acceptance of the scientific community that the indicator that you use represents the concept that you are trying to use. Factor - One of many causes for variation in a specific variable. Frequency distribution - General Social Survey (GSS) - A survey distributed to about 1500-2000 U.S. adult citizens almost every year which asks various questions about attitudes, personal characteristics, etc. Group means substitution - A method of replacing data where you take the mean of each group you are studying and replace each missing case with the value of their group. Hidden populations - People who are difficult to study, often without telephones, addresses, or ways to contact them. Hypothesis - A statement of your research question. Must have two variables linked by a causal statement and should be logically related to your research question and previous research and theory. Hypothesis testing - The steps taken in evaluating your hypothesis. Includes the necessity to throw out hypothesis which do not meet the necessary qualities of causation (temporal order, association, and eliminating alternatives and the need to reject the null hypothesis. Inclusiveness - Inclusive responses to questions ensure that the entire range of responses is available. Independent variable - The variable that is not based on other variables in your analysis. The independent variable is what you want to change to observe changes in the dependent variable. Indicator - The measure that you use to get data on the concept that you wish to study. Indirect relationship - A relationship which you have found that is actually liked together by another variable rather than the specific ones you are looking at. Inferential statistics - A type of statistic that evaluates whether your results from your sample can be applied to the general population. Informed consent - Getting your subjects to sign a form which describes what the research project will entail, how long it will take, and that they are free to refuse to participate at any time. Inter-coder reliability - In content analysis research with multiple coders, it is important to ensure that the coders are interpreting the material in the same way. Inter-coder reliability uses statistical measures to determine if there is enough similarity between coders. Institutional review board (IRB) - Interaction - The effect that questions close in proximity have on each other. Previous questions effect responses on questions which are after them. Interval-level variable - Variables which are measured in standard, equally spaced units without any true zero. Internal validity - Measures how well the measures that you use represent the concepts that you are trying to study. Intervening variable - A variable that comes in between two variables to better explain a relationship between them. Latent coding - When you decide to read into the meaning of the content that you are analyzing to get your data rather than simply taking it at face value. Leading question - A bad type of question where you incline the respondent to answer a certain response through the wording of a question. Intentionally or unintentionally, leading questions can be very problematic for your data. Level of measurement - The how the data which you are collecting is ordered. Can be nominal, ordinal, interval, or ratio. Effects what type of statistics you can use on the data. Likert scale - A type of question which has a range of agreement and disagreement levels in reference to a statement. Listwise deletion - A method of replacing missing data where you simply delete the cases or variables which have any missing data. Longitudinal research - Research which is done in a series of time periods with the same group of people. The same group of people are surveyed at various times to track changes in those people. Manifest coding - Used in content analysis, coding content based on the face-value rather than looking into the meaning. Mean Means substitution - A way of dealing with missing data where you take the mean of the rest of the data for a certain variable and replace the missing data with those values. Measure of association - Statistics which determine relationships between variables. Different statistical measures are used for the different levels of measurement. Measurement validity - How well your measures (questions, codes, etc.) represent the concept which you are trying to study. Median Missing data - Data that you do not have for various reasons. Most common types of missing data are failure to respond to a question, failure of the interviewer to gather the data, and questions which some respondents are not supposed to answer because of filter questions. Missing data bias - Errors in you data which are caused because of missing data. Usually caused by a certain type of person not responding to a question, but can be caused by problems in replacing missing data, deleting missing data, or properties of the cases which you do have data for. Mode Model - A theoretical representation of how some relationship works. Multiple choice - A type of question that has set responses that the respondent must choose between. Multiple dichotomy coding - Takes each possible response and treats it as a separate variable. Each possible response is treated as its own “yes or no” question for data entry. This allows all the data to remain fully in-tact for analysis. Multiple indicators - Two or more questions that get data about the same concept. Multiple response coding - The various methods of coding questions where the respondent can give as many responses as apply to them. The methods include the multiple dichotomy method, the multiple response method, and the frequency method. Multiple response questions - Questions which allow the respondent to select more than one answer to a question. Multiple stage sampling - A method of sample where you take one sample from the population and then take smaller samples from that original sample. Negative relationship - An association between two variables where when one variable increases the other decreases. No harm to participants - A concept related to ethics that requires that participants are not knowing harmed by the research. Nominal-level variable - A variable which is measured by simple classifications. Nominal data cannot be compared numerically, the categories are simply different. Non-probability sample - Samples which do not include an equal chance for everyone in your target population to be selected. Types include convenience, snowball, purposive, and quota. Non-reactive research - Research methods which look at data created without the knowledge of the participants. Avoids reactivity of data. Includes content analysis and non-reactive observation. Non-response bias - Bias due to the type of people who refuse to answer particular questions. If one group of people refuse to answer a question, data for that question will not be representative because your effective sample for that question is no longer representative. Null hypothesis - The alternative hypothesis to your causal hypothesis which states that there is no relationship between the variables which you are trying to link. Open-ended question - A question where you leave room for the respondent to form their own response with no prompts. Operationalization - The process of taking your concept and forming measures to collect data about that concept. Ordinal-level variable - A variable which is measured by ranked categories without any specified distance between the categories. Pairwise deletion - Uses statistical analysis to predict values for the missing data. Partially open question - A question that allows for an "other" category that has space for the respondent to write in their specific response. Population - The entire body that you wish to generalize to and your sample is drawn from. Everything in the population must have an equal chance of being selected in the sample for your sample to be random. Positive relationship - An association between two variables where when one variable increases the other also increases. Prestige bias - A problem with data caused by questions that associate specific answers with a well known figure or group. The association causes respondents to tend to answer like the well-known figure or group. Pre-test (pilot) - A test of your measures (questionnaire, questions, coding, etc.) to make sure that they adequately represent the concept which you wish to study, are not confusing, take the appropriate amount of time, etc. Privacy - A part of ethical research, privacy makes sure that you do not violate the subjects' right to privacy when asking questions. It is also important to keep their contact information private. Probability sample - A sample where the population which you want to generalize to has an equal chance of being selected. Different methods of random sampling include systematic random sampling, simple random sampling, and random digit dialing. Probe - A comment or question made by the interviewer to get the respondent to explain further on a specific topic. Purposive sample - A method of sampling that only includes certain types of people in the sample rather than trying to be random. Works well with small and hard to find populations. Questionnaire - A written collection of questions which you compile, assort, and distribute to your sample to get the data you wish to analyze. Quota sample - A non-probability sample technique which makes sure that certain numbers of people of different groups are included in your sample. Works well to ensure that small groups are included in your sample. Random digit dialing - A random sample technique available when doing telephone interviews where you let computers randomly dial numbers within a specifies area code and local exchange. Random number table - For use when you are using simple random sampling, a random number table is a computer generated table with no pattern to the numbers that allows you to assign a number to each member of your population and use the table to randomly select your sample. Random sample - A sample where each person in the population who you wish to generalize to has an equal chance of being selected in your sample. Random samples allow you to generalize your analysis to the entire population. Ranking question - A question which asks the respondent to analyze multiple choices in relationship to each other rather than independently. Ratio-level measurement - Categories of data which have a specified order with equal distance between the categories and have a true zero as a base. Rating question - A question which asks people to assign a numerical (or other) response (out of 5, 7, 10) to items in a question. Reactive data - Research methods where the subject knows that they are being researched produce reactive data which is somewhat altered because of the process of conducting the research. Record - The collection of the coded data for a case. Reliability - How well the meaning of questions and answers remain the same regardless of who is reading them. Replication - An important concept in scientific research, replication is the opportunity to check the results of a research project. Transparency of the method used is important for exact replication. Representative - How similar your sample is to the general population. Response rate - The number of people who answer your survey compared to how many were in your sample. Sample - The people who you select to survey out of your population. Sampling error - Problems in your data due to bad sampling technique and execution which cause differences in your sample and the population. Sampling fraction - The ratio of number of people in your final sample to the number of people in the population. Sampling frame - The people who have the potential to be selected in your sample. Scale - A combination of similar individual measures which are coded to form one measure for analysis. Secondary analysis - A research method that takes data which has already been collected for other purposes and applies it to you specific research question. Semantic differential - A type of question which sets up two opposite concepts, attitudes, etc, and asks the respondents to choose which is closer to their own opinion or which describes themselves or others best. Significance tests Simple random sample - A random sampling technique where you assign a number to each person in the population and then use a random number list to select the number of people who you want in your sample. Snowball sample - A non-random sample used primarily for small or hidden populations where you take the people who fit your criteria and use them as contacts for other people who also fit the criteria of your study. Social desirability - A problem with data which can occur because of people's wish to answer in a way that they believe is socially acceptable rather than their own opinions. Specification - Spurious relationship - An association between two variables which can better be explained by another variable. Stability reliability - A consistency of meaning over time. Administering the survey to the same people on different occasions is the best way to test. Stratified sample - A random sample based first on categories of people who you wish to study. Categories are selected and then other random sampling techniques are used to select your sample from those categories. Subject - The people or artifacts that are included in your sample and are actually included in you research. Systematic sample - A random sampling technique where you find your sampling ratio and go through a list of your population and select every x entry in the list, where x is your sampling ratio. Target population - The people who you want to generalize your study to and your sample is drawn from. Your target population must all have an equal chance to be included in your sample. Tautology - A logic error based on circular reasoning, meaning that something is true be definition. Your dependent variable is simply a restatement of your independent variable Teleology - A logic error which explains a phenomenon by saying that it was some spirit or higher power that causes the relationship. Temporal order - One of the requirements of causation which specifies that the causing variable comes before the variable changing as the effect. Theory - An possible explanation of how a variable or set of variables works. Theory testing - A type of research which seeks to prove or disprove a theory through doing further research on the topic. Takes theoretical expectations and tests them against empirical observations. Unit of analysis - What cases or units you are studying, especially relevant in content analysis where you could study words, phrases, advertisements, etc. Validity - How well your measures represent the concept that they are attempting to measure. Variable - An aspect of something that changes or has more than one category. Variance - The amount of change in the answers to a question. Voluntary consent - A part of ethics, voluntary consent makes sure that the subjects are not coerced into being part of your project.
|