Meaning: A specific type of experiment where people are randomly shown one of two versions of something to compare how each of them performs.
It is typically used to refer to a type of experiment that does not include a control condition, though this terminology is not universal.
Example: Someone wants to examine which of two messages is more effective in getting people to read more about farmed animal welfare, or which donation appeal leads to more donations received.
Meaning: Usually, when people talk about a hypothesis, they are actually referring to the alternative hypothesis. It is a prediction or educated guess about a study result like the nature of a relationship between variables. The opposite is the null hypothesis, in which there is no significant relationship between those variables.
Example: You’re interested in seeing if reading a leaflet about factory farming reduces meat consumption. Your alternative hypothesis would be that there is a significant difference in meat consumption between people who read the leaflet and people who do not.
Meaning: When participants drop out of a study partway through, that’s attrition. For instance, after completing half of the questions on a survey, or participating for two months out of six. Differential attrition is much more of a problem than random attrition.
Example: In our Going Veg*n study, we asked people transitioning to veganism and vegetarianism to complete a series of surveys over a six-month period. Fewer people completed the follow-up surveys as the study progressed. In the end, 65% of the initial participants completed the entire study, meaning there was 35% attrition.
Meaning: Bias in a study refers to a problem in the way you collected the data that will make the results look different than they should. A well-designed study minimizes the chance of bias. This is different from error.
Example: In a survey about plant-based products, if you include too many vegans, you’re likely to bias your results in an overly positive direction (because vegans are probably more positive about plant-based products than non-vegans).
Meaning: Stands for “Black, Indigenous, and People of the Global Majority.” This term describes the majority of the world’s population, which is non-white (yet often referred to as “minorities”), and recognizes the unique struggles suffered by Black and Indigenous people.
Example: Our study on the relative effectiveness of different advocacy tactics looked at pro-animal behaviors and attitudes of different racial/ethnic groups. We suspect that more speciesist beliefs observed among African Americans can likely be attributed to structural inequalities and their continued marginalization, while also recognizing that animal advocacy research pertaining to BIPGM experiences is scarce. As a result, we encourage cultural research to better understand these issues.
Meaning: Stands for “Black, Indigenous, and People of Color.” By naming Black and Indigenous peoples, this term recognizes the violence and struggles faced by these groups in particular.
While BIPOC is a widely used term, we often use the term BIPGM, as it explicitly decentralizes whiteness by noting that non-white races are the global majority.
Example: Our Going Vegan or Vegetarian study found that there was no clear association between success on a new veg*n diet and being white or BIPOC.
Meaning:A detailed investigation of a specific phenomenon, usually one type of industry, group of people, or animals. Case studies can employ qualitative or quantitative methods.
Example: We conducted a case study on the number of vegetarian and vegan purchases from a single restaurant to track whether purchases changed over time.
Meaning: Causality refers to a relationship between two variables where we can show that one causes differences in the other (cause and effect). It is important to note that a correlation between two variables does not imply causality. Experiments with a control group are best for determining causality. Causality is sometimes referred to as causal direction or causation.
Example: Faunalytics looked at the impact of different animal advocacy tactics on animal product consumption and used an experiment to figure out causality. Using an experiment let us be sure that advocacy experiences were causing the behavioral and attitudinal changes we observed.
(Also seen as 𝛸2 test. Pronounced “kai square,” where “Chi” rhymes with “eye.”)
Meaning: A statistical test that tells you whether there is a statistically significant association between two sets of categories (such as gender, race/ethnicity, education level, or type of diet). That is, do people who differ on one tend to differ on the other?
Example: Our study of Beliefs about Fishes and Chickens used a chi-square test to determine if there were significant differences in the rates of men versus women (one set of categories) who signed versus didn’t sign a petition (another set of categories).
Meaning: Coding is the process of reading through open-ended data to develop themes for the thematic analysis of the data. It involves categorizing data (usually written text or transcribed audio) into chunks that describe what the data is expressing. This is a way of simplifying open-ended data so that we can analyze it better.
Example: To analyze why former veg*ns abandoned their veg*n diets, we coded their responses to open-ended questions. During the coding process, we grouped participants’ written reasons for abandoning their veg*n diets into themes: unsatisfied with food, health, social issues, inconvenience, cost, lack of motivation, and other. Under the theme of being unsatisfied with veg*n food, for instance, people wrote things about cravings, being hungry, wanting more variety, and so on.
Meaning: Conditions are what we call the different experiences people are given in an experiment. Participants in an experiment are randomly assigned to one of several conditions, which will be as similar as possible except for one element (the manipulated variable). With that setup, the researcher knows that if people in different conditions respond differently, the only possible explanation is that the manipulated variable caused the difference.
Example: In our “Reduce” or “Go Veg” study, participants were assigned to one of three conditions: a vegetarian advocacy video, a reduction advocacy video, or a control video.
Meaning: This indicates the precision of the estimates we get from a study. A study’s sample is just a slice of the overall population, so a confidence interval (CI) is used to indicate the highest and lowest plausible population value. A smaller CI means we can be more precise about the estimate. Most studies report 95% confidence intervals, which means that if we repeated the study 100 times, we would expect the true population value to fall within that interval 95 times.
Example: In our first Going Veg report, Figure 1 shows that 66.9% of vegetarians were omnivores before going veg. In parentheses, it then says +/- 10, meaning that the CI on that estimate is 56.9% to 76.9%—sometimes written as (56.9, 76.9). In other words, while 66.9% is our best guess of what we’d find if we could survey every member of the population, it could plausibly be anywhere between 56.9% and 76.9%.
Meaning: Participants in this condition of a randomized control trial do not receive an experimental manipulation.
Example: In our experiment on the effectiveness of video outreach on pork consumption, the control group didn’t watch a video on factory farming while the treatment group did.
Meaning: A measure of the relationship between two variables. A positive correlation means that values of both variables go hand in hand so that a higher value of one tends to go with a higher value of the other. A negative correlation is when lower values of one variable are related to higher values of another variable.
Example: In our study of beliefs about chickens and fishes, we found a positive correlation between the belief that fish have a personality and taking a pledge to reduce one’s consumption of fish, meaning that individuals were more likely to take a diet pledge if they also held stronger beliefs that fish have a personality. We also found a negative correlation between the belief that chickens have no personality and signing a petition to improve their welfare, meaning that individuals were less likely to sign a petition if they also held stronger beliefs that chickens had no personality.
Meaning: A study conducted at a single time point that compares two or more groups. This is often contrasted against a longitudinal study, where a single group of participants is tracked over time.
Example: A lot of our studies collect data from participants only once. For instance, our landmark study of Current and Former Vegetarians surveyed participants once, and compared people who currently or previously ate a vegetarian or vegan diet to learn about what makes people continue or quit.
In contrast, our longitudinal study involved multiple surveys of a single group of participants who were newly vegetarian or vegan. In this case, we surveyed them monthly for six months to learn about what made them continue or quit.
Meaning: The main variable that is being measured or tested in a study. Its values are assumed to be influenced by the independent variables (i.e., their values ‘depend’ on the independent variables).
Example: We often measure how behaviors or attitudes, the dependent variables, are affected by an independent variable in our studies. Some dependent variables we look at a lot are self-reported animal product consumption, or whether or not someone signs a welfare petition.
Meaning: They help describe and summarize a dataset. These can include things like the mean, median, standard deviation, and variance of your data. Descriptive statistics can also include graphs that show you how your data are distributed, such as in a bar chart and graphs that explore the relationships between multiple variables.
Example: Our report about the people who support animal causes includes descriptive statistics like the range of donations to animal charities ($1 to $10,000) and the amount donated to animal charities by the median donor ($90). A graph provided in the study also shows the distribution of all the donation amounts, indicating what percentage of people donated a specific amount of money to animal charities.
Meaning: An inexpensive and relatively quick research method in which existing data and information are summarized and analyzed for a different purpose than the original source. Sources of information can include articles, books, government databases, published reports, websites, and more.
Example: Faunalytics’ study to understand the industry costs for chicken, egg, and fish products in the U.S., Brazil, and China involved conducting desk research. This meant reviewing industry, economic, and academic analyses and data related to the target industries and countries.
Meaning: See attrition first. Differential attrition is the problematic kind, when the people who dropped out of your study were different in some way from those who stayed in. This can introduce bias into your results.
Example: We looked for evidence of differential attrition in our six-month study of new vegans and vegetarians by examining the characteristics of the people who left the study in comparison to those who remained. 35% of the sample dropped out over the six months, so it could have been bad if those people were different in some way from the ones who stayed, as this would have implied differential attrition.
Meaning: Refers to the differences that make people unique, like race, ethnicity, gender identity, sexual orientation, age, (dis)ability, religion, socioeconomic status, political affiliation, national origin, language, culture, and more.
Example: As we note in our study on the relative effectiveness of different types of advocacy, it’s important to recognize the diversity of communities that advocates work with and how marginalized groups are often underrepresented in research. As a result, we looked at whether the effects of animal advocacy varied among racial/ethnic groups and what impact race/ethnicity had on the study’s outcome measures, so that we could identify opportunities and challenges.
Meaning: This is a measure of the strength of a relationship between two variables, or the difference between two groups. These include correlation and regression coefficients, and r2. Regardless of the type of effect size and whether it is negative or positive, larger numbers indicate a stronger relationship between variables or a bigger difference between groups.
Example: In our study on U.S. beliefs about chickens and fishes, we found the strongest correlation (the largest effect size) between beliefs about fish personalities and taking a diet pledge, meaning that there is a strong relationship between believing that fish have personality traits and taking a diet pledge. On the other hand, there was a weak correlation (small effect size) between beliefs relating to fish as food and signing a diet pledge, meaning that there was a weak relationship between the two.
Meaning: Equality refers to treating all people the same, so as to promote fairness. It does not consider the different experiences lived by people and the structural inequalities that provide an advantage to some and a disadvantage to others.
Example: Let’s say two people from different economic backgrounds want to care for companion animals. Treating these two people equally would mean that both people are given the same support to care for their companion animals, such as providing both with a leash and dog food, regardless of whether or not they require additional assistance or no assistance at all. Unlike equity, this approach doesn’t take into consideration people’s varying needs based on their circumstances. Treating people equally rather than equitably can perpetuate harms against already marginalized groups.
Meaning: Unlike equality, equity takes into consideration the diversity of people and their different lived experiences. This means understanding what different groups of people need in order to be successful and adjusting policies, systems, attitudes, and resources accordingly.
Example: Let’s say we have the same two people from our equality example above, who both want to adopt companion animals. To treat them equitably we would need to ensure that both people have the same ability to care for their companion animals by providing each with the materials and assistance they need to do so. This may mean providing financial assistance or supplies like dog food to one person, while the other person may require less assistance.
Meaning: A type of study where participants are randomly assigned to one of 2+ conditions to investigate whether a particular experience causes a particular outcome. Experiments with control conditions are sometimes called randomized control trials.
Example: To determine whether particular types of advocacy change behavior, we conducted an experiment as part of our study on the relative effectiveness of different advocacy tactics, in which participants were randomly assigned to view one of ten forms of animal advocacy or to a control condition.
Meaning: Statistical analyses are probability-based: they tell us how likely it is that a difference we see in a sample is highly likely to be true of the entire population. The problem is that the more analyses you conduct, the more you increase the chances that your effect will be statistically significant when there really isn’t a real effect in the population—that is, a false discovery. We make corrections to our statistical tests to prevent the False Discovery Rate from affecting our results so strongly.
Example: In our study on the labeling of plant-based meat alternatives, we ran comparisons of eight label options using t-tests that were corrected for FDR because of the many comparisons involved. (With eight labels, there could be up to 28 comparisons between pairs of labels.)
Meaning: A type of study similar to an interview, but conducted in a group setting rather than with an individual person. Focus groups are used to gather information about a specific topic, for which multiple people contribute to answering a set of questions. They provide rich and detailed information on a subject, but because of the small number of participants, we can’t assume that the frequencies of participants’ opinions will generalize to the entire population.
Example: You’re interested in getting detailed information about why people adopt companion animals and the experiences they have when adopting, so you decide to conduct a focus group with 8 people who adopted companion animals. Thematic analysis of the focus group’s answers can help you interpret what you found or even design a survey to get data from a larger sample.
Meaning: How well the results and conclusions from a study conducted on a sample population can be applied to the larger population. Typically, the larger the sample size, the more generalizable the results (unless the sample was collected in a way that could introduce bias).
Example: Our results regarding the impact of corporate commitments to switch to cage-free eggs on public attitudes are not generalizable to all potential egg consumers in the U.S. because the sample for the study came from Facebook users who chose to interact with articles about cage-free commitments, which does not capture the general U.S. population. Instead, we might expect it to generalize to all people who read news on Facebook.
Meaning: Mental “shortcuts” that people rely on when making decisions, especially when faced with limited information. Common heuristics include the availability heuristic (making a judgment based on limited, existing information), the anchoring and adjustment heuristic (forming an initial judgment and then seeking out future information to justify this opinion), and the representative heuristic (judging the probability of an event based on a stereotype, average, or existing opinion).
Example: You come to the conclusion that all plant-based meat substitutes taste unnatural after trying only one product.
Meaning: An educated guess or prediction that can be tested through an experiment or through observation. Hypotheses are often expressed in an “if” and “then” statement.
Example: You hypothesize that reading a leaflet about factory farming will reduce meat consumption. Expressed in an if/then statement, this would be: “If meat consumers read a leaflet about factory farming, then they will reduce their meat consumption.”
Meaning: A method for testing if the results of a study support the researcher’s predictions or not. It requires that the hypothesis be expressed as a null hypothesis and an alternative hypothesis so that the statistical results tell you whether you can reject your null hypothesis in favor of your prediction.
Example: You want to test your hypothesis that reading a leaflet about factory farming will reduce meat consumption, so you decide to conduct a t-test to see whether there is a statistically significant difference in meat consumption between people who read the leaflet and people who don’t.
Meaning: Inclusion means including, welcoming, and valuing the diversity brought by different groups of people and their lived experiences, particularly those who have often been excluded throughout history.
Example: In our Going Veg study we found that about half of vegan participants could be considered “plant-based eaters” rather than “lifestyle vegans,” since their avoidance of animal products was mainly limited to food. However, we did not differentiate between vegans and plant-based eaters in this study, in part because our goal is to support all people who are working to reduce harm to animals.
Meaning: When conducting research, it is important to consider what characteristics you want your sample of participants to have in order to best answer your research question. These are often demographic characteristics, such as geographic location, gender, and diet type.
Example: Our study of people transitioning to veg*nism required that participants be adults in the U.S. or Canada who had started a new vegan or vegetarian diet within the past two months. These were the inclusion criteria for the study.
Meaning: Qualitative research often uses interviews to gather information from people on a specific topic, by asking people questions in a conversation.
Interviews can follow a list of questions (called structured interviews) or they can be unstructured, where the interviewer bases questions off the participant’s previous responses. Semi-structured interviews are a middle ground, and are the type Faunalytics most often uses.
Example: We conducted interviews with individuals who completed a farm sanctuary tour to better understand their experience and how it affected them.
Meaning: A variable that is being tested to see whether its presence or specific characteristics influence the dependent variable.
Example: In our study on the relative effectiveness of different advocacy tactics, we conducted an experiment in which our independent variables included different forms of animal advocacy (such as social media posts, graphic and non-graphic videos, leaflets, billboards, and more). We analyzed how effective each tactic was at reducing people’s consumption of animal products (the dependent variable).
Meaning: Inferential statistics go a step beyond descriptive statistics by using data to look for a systematic association between variables that can be generalized to the population the sample came from.
Example: Our experiment on whether messages and donor characteristics can increase donations to farmed and companion animals analyzed whether there were significant differences in donations among four types of appeals, between companion animals and farmed animals, and between people who had previously donated to animal charities and those who hadn’t. These analyses allowed us to make predictions about donations to farmed animal and companion animal charities in general.
Meaning: A way of measuring people’s emotions, beliefs, intentions, or attitudes that gives a numerical value to the strength of their opinion. These scales are usually labeled so that, for instance, 1 = ’strongly disagree’ and 7 = ‘strongly agree,’ with 4 as the neutral midpoint.
Example: Many Faunalytics surveys use Likert scales to understand people’s attitudes, beliefs, intentions, or emotions. For example, in our study of the impact of different animal advocacy tactics, we measured people’s reactions towards the advocacy on a 7-point Likert scale from ‘strongly disagree’ to ‘strongly agree’.
Meaning: No study is ever perfect. Some factors are out of the researchers’ control, which could affect the interpretation or scope of results. These factors are referred to as limitations and should be listed in the report.
Example: You’ll find limitations of our original studies listed towards the end of our reports. For example, in our study on the state of animal advocacy in the U.S. and Canada, one limitation was the small number of marginalized advocates that were part of the study. This meant that all historically marginalized groups (BIPGM, people with disabilities, and people of the LGBTQIA+ community) had to be combined into one group for our analyses, which may have washed out differences among the groups.
Meaning: A study that involves repeated observations or testing of participants over a period of time.
Example: In our study looking at people who just started a vegetarian or vegan diet, we collected data from these individuals across six months to learn about what made them continue or quit.
In contrast, our 2014 study of Current and Former Vegetarians was cross-sectional: We surveyed participants once, and compared people who currently or previously ate a vegetarian or vegan diet to learn about what makes people continue or quit.
Meaning: When a researcher ‘manipulates’ or controls the independent variable, it refers to changing just one element of an experience in a controlled setting to observe its effect.
Example: In our study of the impact of different animal advocacy tactics, we manipulated the type of animal advocacy tactic (e.g. videos, leaflets, etc.).
Meaning: This shows you how representative or accurate your results are to the target population. The smaller the margin of error, the more your results reflect the overall population. Typically, the margin of error is smaller (better) when the sample size is bigger. For a more detailed explanation of how we calculate margin of error, check out the FAQs section of our Research Advice page.
Example: In our study about people who support animal causes, we estimated that approximately 5.7% of study participants that had donated to animal causes were veg*n, with a margin of error of ± 1.4%. This means that we can be confident that in the general population of people who donate to animal causes, 4.3% – 7.1% are veg*n. These results suggest that the rate of veg*nism among people who donate to animal causes is higher than in the general population.
Meaning: This essentially means that although we didn’t find a statistically significant result, it was close enough to consider potentially meaningful, just with less certainty. While a p-value under .05 is typically considered statistically significant because it indicates 95% confidence in the result, a p-value a little higher than .05 but less than .10 is sometimes referred to as marginally significant because it still indicates 90% confidence in the result.
Example: In our study of new veg*ns (Table 4), we found that people pursuing a vegetarian diet were marginally more likely to have tried a fad diet than people pursuing a vegan diet. The difference between vegans and vegetarians was marginal with a p-value of .06.
Meaning: Also known as the average, this is a handy measure to understand a variable’s overall value or score. It’s a single value that is obtained by adding all the values of a single variable, and then dividing the sum by the number of values.
Example: In our study about the effects of farm sanctuary tours on people’s intentions and diet change, we asked participants if they believe that eating animals directly contributes to the suffering of farmed animals. Responses were made on a Likert scale ranging from 1, not at all, to 4, strongly. We found a mean of 3.33 in the pre-tour group and a mean of 3.60 in the post-tour group. This was a statistically significant difference, showing a shift in attitudes.
Meaning: Similar to the mean, but this is just the middle value of all observed values (no math required). Half the values are above this midpoint and half are below it. This can be a more useful measure than the mean in cases where some of the numbers are much higher or lower than most of them.
Example: Start by ordering the numbers from smallest to largest. Let’s say we have the numbers 2, 4, 6, 7, 1227. The middle value here is 6, so that’s the median. If we had used the mean instead, it would be much higher because of the 1227, which would give a misleading impression of the data.
Meaning: A type of quantitative study that combines the numeric results from multiple published studies on a similar topic to get an overall impression of the topic. This is a useful way to summarize current literature.
Example: Researchers conducted a meta-analysis that looked at the overall effect of animal welfare messaging on people’s consumption of animal products. The meta-analysis found that interventions in 71% of the analyzed studies succeeded in reducing animal product consumption, at least in the short-term. Based on the results from the meta-analysis, the authors were able to make recommendations to help future researchers provide more reliable information about what really works to influence meat consumption.
Meaning: When there is a systematic difference between individuals who participate in a study and those who do not. These differences then ‘bias’ the results since they are only representing the group of individuals who agreed to participate and not the entire population.
Example: The results from the follow-up survey for our study on a farm sanctuary tour’s effects on intentions and diet change tended to come from people who were already consuming fewer animal products than those who didn’t respond. This means that the results from the follow-up survey should be interpreted with caution.
Meaning: Predictions in a study are usually about a significant relationship between variables or a difference between the groups tested. The null hypothesis is the opposite: that there is no significant relationship or difference.
Example: If one tested the hypothesis that reading a leaflet about factory farming reduces meat consumption compared to a control group, then the null hypothesis would be that there is no significant difference in meat consumption between the groups tested.
Meaning: Surveys sometimes include open-ended questions, where participants write about their thoughts or feelings about a topic in a couple of sentences (or more).
Example: Faunalytics conducted a poll shortly after the COVID pandemic began, to better understand the public’s knowledge of the relationship of the pandemic to animals. One of the questions was open-ended, asking people to explain their understanding of where the disease came from.
Meaning: This is a small study that is conducted before the intended study to test its feasibility or select the methods that will be implemented. Pilot studies allow us to make changes and improve the study design before officially running the study.
Example: Before conducting our large-scale field study with Animal Equality on the effectiveness of Animal Equality’s video outreach, we ran three pilot tests. Running these pilots allowed us to fine-tune the way we were measuring participant diets until we felt comfortable with the measure to proceed with the large-scale study. By conducting the pilots, we were able to address a variety of issues before beginning the more time- and resource-intensive main study.
Meaning: P-values are often used when you conduct statistical tests on a sample rather than an entire population. A p-value after a statistical test tells you how likely it is that you would see a difference that big or a correlation that strong in your sample if there’s no actual difference or correlation in the population. Basically, it tells you how likely it is that you’re wrong when you say that there is a difference or a correlation.
P-values have predetermined thresholds that allow us to determine if a result is statistically significant or not. This value is often 0.05, meaning there is less than a 5% chance that you would see those results due to random chance.
Example: In our study of animal advocate experiences and turnover, we found that paid advocates in our sample were significantly more likely to experience burnout than unpaid advocates (p < .01). P-values correspond to percentages so this means that the chance that there is no difference between paid and unpaid advocates in the full population of advocates is less than 1%.
Sometimes we will refer to several results at once if their p-values are all in the same range. In this case you will see the plural ps. For instance, in Table 5 of the same report, you will see a footnote “*Paid and unpaid advocates significantly differed (ps < .05),” which indicates that the p-values (ps) for all results marked with an asterisk were less than .05, meaning they were statistically significant.
Meaning: This is generally a short survey, often just one multiple-choice question, used to obtain quick information on people’s opinions, attitudes, preferences, and demographics.
Example: Shortly after the start of the COVID-19 pandemic, Faunalytics conducted a five-question poll to see what the general public knew about the pandemic’s relationship to animals.
Meaning: A group of people that can be defined by one or more shared characteristics. Although the word is most commonly used to refer to people who live in a particular country, that is only one definition.
Example: Our study of advocate retention was completed by participants from the population of all animal advocates in the U.S. and Canada.
Meaning: An analysis that is run before collecting data to determine the smallest sample size needed to detect a difference or relationship between the variables, if there is one.
Example: Before conducting our study on plant-based meat alternative labeling, we ran a power analysis to determine how many people would need to be included in the study. The power analysis showed that we would require 1340 people to conduct our analyses. Because our sample size was even greater than that, our analyses had more than 99% power to detect effects, so we can be very confident that the results would be the same if the study were run again.
Meaning: Preregistration is a way for a proposed research plan and a study’s hypotheses to be defined before beginning data collection and obtaining any results. You may have heard someone say that statistics can be made to show anything you want. Though exaggerated, there is a grain of truth to this because it’s usually possible to analyze the same data in different ways, which may provide different results. Preregistering the analysis method provides proof that the researchers didn’t try dozens or hundreds of different methods until they found a result that they “liked.”
Example: All of Faunalytics’ original studies are preregistered on the Open Science Framework. There, we provide a description of the study we will conduct, we list our research questions and hypotheses, describe the data collection process, and our proposed analysis.
Meaning: Research that doesn’t involve measuring things on a numerical scale. Instead, qualitative data often includes people’s written or oral statements, collected through interviews or open-ended questions on a survey.
Example: Faunalytics’ study about how to support farmed animal protection in China involved conducting interviews with members of the Chinese farmed animal protection community, which were then analyzed for common themes.
Meaning: Research that provides numeric data such as percentages and averages.
Example: Our studies on attitudes towards chickens and fishes in the United States, Brazil, Canada, China, and India provide numeric information like the frequencies of key beliefs about chickens and fishes, as well as correlations showing how they’re linked to signing welfare petitions and taking diet pledges to reduce consumption.
Meaning: See regression first. This is a value that’s associated with regression models and tells you what percentage of the variance in the dependent variable is explained by the independent variable.
Example: In our first Going Veg*n report, we examined R2 to determine whether certain demographics were meaningful predictors of consumption success on a veg*n diet. We used R2 = 0.04 as our minimum cutoff, which meant that at least 4% of the variation of the dependent variable (consumption success) was explained by the independent variables in the regression model (age, romantic status, number of children).
Meaning: While attrition, or participant drop-out, can be a problem in research, random attrition is much less concerning than differential attrition. That’s because when random attrition occurs, the participants who drop out are similar to those who stay, so the chance of bias is low.
Example: In our six-month study of new vegans and vegetarians, we examined how many participants dropped out over time. There were no significant differences in the characteristics of people who completed the study and those who didn’t, so this was likely random attrition.
Meaning: The term random error (often just called “error”) refers to random differences between individual participants and the average of all participants. It’s important to understand the difference between error, which is random, and bias, which is systematic. Bias is a problem for a study, while random error is unavoidable and the reason we use statistical tests to interpret our data.
Example: Imagine we survey people’s attitudes about plant-based products and the average (or mean) attitude score turns out to be 4.2 on a 7-point scale, around the middle. Some people’s attitudes are higher and some are lower than 4.2. Those differences are collectively referred to as error as long as there is nothing systematic about them. If we were to find that women’s attitudes tend to be more positive than men’s and we had 80% women in our study, then there may be a systematic bias in the study instead of just random error: attitudes may appear to be more positive in this sample than they really are in the population.
Meaning: This is a particular type of experiment. In experiments, participants are randomly assigned to one a treatment group or a control condition, which is where the “randomized” in the name comes from. In this type of experiment there is always a control condition, which is where the “controlled” part comes from.
Example: In the second study in our Planting Seeds report, participants were randomly assigned to experience animal advocacy or a control condition. Although we refer to it as an experiment in the report, the features of randomization and a control condition mean that this study could also be called an RCT.
Meaning: A type of statistical analysis to look for a relationship between one or more independent (predictor) variables and a dependent (outcome) variable.
Example: In our first Going Veg*n report, we conducted a series of regression analyses to see whether things like participants’ transition speed and level of commitment to their diet could predict their success on that diet.
Meaning: See regression first. The regression coefficient indicates the relationship between the dependent variable and an independent variable. The value of the coefficient is the amount by which the dependent variable increases (if the coefficient is positive) or decreases (if the coefficient is negative) for every unit increase of the independent variable.
Example: In our study of the obstacles that people face when they pursue a veg*n diet we ran a regression analysis to see which demographics predict how many monthly servings of animal products people eat after lapsing. The regression coefficient associated with having a veg*n-positive attitude post-diet was -9.7, meaning that people who expressed veg*n-positive sentiments ate about 9.7 fewer servings of animal products per month than people who didn’t.
Meaning: This refers to the consistency of a measure – basically, how similar your results would be if you were to repeat the measurement under the same conditions over and over again. A reliable measure should always give you similar scores if you’re measuring from the same population under the same conditions.
Example: Measures of speciesism were shown to be reliable after showing consistent results over a four-week period. Keep in mind that while a measure can be reliable, it isn’t necessarily valid: for instance, measuring shoe size as an indicator of speciesism would be reliable because shoe size doesn’t change much, but it certainly wouldn’t be valid. A good measure should be both reliable and valid!
Meaning: No study can ever definitively tell you what is happening in a population, because there is always a degree of chance involved, and decisions about how to measure things. Replication means repeating a study, with either a direct or conceptual replication. Direct replication is as similar as possible to the original study, conducted the same way, under the same or very similar conditions, which helps us be sure the results weren’t due to chance. A conceptual replication tests the same hypothesis but uses a different method and/or measurement tools. This helps us see whether the results were due to specific choice of methods or measures rather than something deeper.
Example: Our second study of Attitudes towards Chickens and Fishes, which was conducted in Brazil, China, India, and Canada, was a partial replication of an earlier study conducted in the U.S. The method was directly replicated (with the exception of translation and a few cultural adjustments) but the study wasn’t a true replication because the population was different.
Meaning: When we talk about a sample or a study being representative, we mean that the people who participated should share important characteristics with the entire population they came from.
Example: Getting a representative sample is more difficult in some countries than others, particularly if a large part of the population doesn’t have internet. For instance, in our study of the Chinese public’s beliefs about COVID-19’s connection to animal agriculture, we had to use a sample of Chinese adults who were demographically similar to Chinese adults with internet access rather than the entire population of China. As noted in that report, Chinese adults with internet tend to be younger, more educated, and more urban than the general Chinese population.
Meaning: This is the question that a research project or study is trying to answer. For help choosing a useful research question in your own work and learning how to frame it, check out our Literature Search and Review page.
Example: In Faunalytics’s study on the state of animal advocacy in the U.S. and Canada, we had four main research questions we sought to answer:
1. Why do animal advocates leave the movement?
2. Why do animal advocates leave organizations?
3. When do animal advocates leave an organization?
4. What experiences do animal advocates have in the movement?
Meaning: A group of participants from the population the study is based on. The sample should be representative of the population it’s from, in order to generalize the study’s results to the overall population. To achieve representativeness, the sample should be selected from the population as randomly as possible—though this often poses practical challenges.
Example: In our study on the beliefs about chickens and fishes in the U.S. we surveyed a sample of U.S. adults that was representative of the adult U.S. population to see how much they agreed or disagreed with different beliefs about chickens and fishes. The results from this study could then be used to represent the beliefs about chickens and fishes held by the general adult U.S. population.
Meanwhile, our study on the state of animal advocacy in the U.S. and Canada intended to find out about the experiences of current and former animal advocates in the two countries. This meant that our sample had to be representative of the population of advocates in the U.S. and Canada.
Meaning: Saturation is a term used in qualitative data collection. It is achieved when collecting more data is no longer providing new information. Essentially, it tells us when we can stop collecting data.
Example: You decide to stop interviewing people when the new people you interview are mostly saying things you’ve already heard from previous participants (you’re no longer getting much new information).
Meaning: Directly asking participants about their feelings, beliefs, attitudes, or behaviors.
Example: Many of our studies involve self-report data, such as asking people about their diet or thoughts about different animal issues. For example, our Going Vegan or Vegetarian study included a great many self-report questions about people’s diets and dietary journey.
Meaning: Sometimes people’s answers may not be fully accurate when we ask them to self-report, either because of memory recall difficulties or because they want to appear as a better version of themselves (i.e., social desirability bias). Observable measures, like behavior, minimize self-report bias.
Example: To avoid the problem of self-report bias, our “Reduce or Go Veg” study used behavioral measures of diet change, directly measuring whether people took a diet pledge and which foods they ate.
Meaning: This is a way of identifying the emotions behind a text, such as in a response to an open-ended survey question. A sentiment analysis can be as basic as determining whether a response is positive, negative, or neutral, or it can be as complex as identifying the emotion behind the response (for example, sadness, anger, happiness, or surprise).
Example: To gain some insight into public perceptions of different animal-friendly diets, our study on Twitter trends used a sentiment analysis to compare the positivity and negativity of tweets containing similar keywords to see whether there were differences in how each keyword is normally used.
Meaning: When participants give responses that they think they should give rather than their true beliefs. These responses often reflect social norms or their idea of what the researcher wants to hear. This bias can be minimized by not revealing too much of the study’s purpose and hypotheses to the participants until the end.
Example: To prevent social desirability bias in our Reduce or Go Veg study, we measured participants’ food choices at a cafe using their receipts so that they wouldn’t know until later that we were looking at whether they chose meat or not. Having them self-report their diet choices right after watching a video encouraging reduction or vegetarianism would have made social desirability bias pretty likely.
Meaning: The belief that one species is more important than another – often that humans are superior to other animals. Speciesism varies from person to person and we can measure it using a scale.
Example: Our study on the relative effectiveness of different advocacy tactics measured speciesism on a 7-point scale from ‘strongly disagree’ to ‘strongly agree’. We found that speciesism was not affected by animal advocacy experiences.
Meaning: A measure of how much variation there is in the values of a variable. Specifically, this tells us how much responses tend to differ (deviate) from the average. The actual calculation is a bit more complicated, but you can think of it as the average amount that people deviate from average. A low value indicates that people are pretty similar on this variable, while a high value indicates a lot of variation between people.
Example: In our longitudinal study on motivations and influences for going veg*n, we defined “high” external motivation (wanting to go veg*n as a result of other people) as being one standard deviation above the mean. In this case, the mean was about 2.6 and the standard deviation was about 0.8, so we defined high external motivation as a score of 3.4 or higher on the 1-to-5 scale.
Meaning: This is similar to standard deviation in that it measures the spread of values, but it is slightly different because rather than just looking at your sample, it gives you an estimate of how far the mean of your sample is from the mean of the population. Standard error is used to calculate confidence intervals. If you see it in a report, it gives you a rough idea of how much uncertainty there is about an estimate: estimates with smaller SEs are more certain than other estimates with higher SEs.
Example: Faunalytics doesn’t tend to report standard errors in our work because 95% confidence intervals are more intuitive, but you may come across it occasionally. Hypothetically, you could see a study where the average donation to companion animals was $101.45 (SE = $14.88), while the average donation to farmed animals was $46.92 (SE = $30.72). This indicates that on average, people donated more to companion animals, but there was about twice as much variability in how much people donated to farmed animals.
Meaning: Outside of research, “significant” usually means important. However, statistical significance does not mean something is important, only that it is unlikely to have occurred by chance. A significant result can be big or small, important or unimportant, but it is unlikely to be a fluke. In other words, if you ran the same study again, you would probably see a similar result—whether that is a difference between two groups, a correlation between two variables, or something else.
Statistical significance is usually indicated by a p-value.
Example: In our study on the effects of a farm sanctuary tour on diet changes, we found that following the tour, people were significantly more likely to say they planned to eat less animal products (p < .05) than before the tour. However, the difference in the number of people who intended to go vegan before versus after the tour was not statistically significant (p > .05).
Meaning: This is a part of the research that is kept separate from the main report. In the case of Faunalytics’ reports, we usually use this for more complex or less relevant results that we don’t want to include in the main report and/or more technical details regarding the methods used in the study.
Example: Our Going Vegan or Vegetarian study has an extensive Supplementary Materials section that details the methods used in the study in a much more technical way than is provided in the main report. This is useful to members of our audience with research training who may want to better understand or even replicate the study.
Meaning: A research method in which people answer questions—for instance, about their attitudes, beliefs, or behaviors. Surveys can be distributed online, in person, or via telephone.
Meaning: This involves summarizing information from existing studies on a particular topic to answer a research question. This means searching academic databases, and sometimes, gray literature (i.e., literature produced outside of the academic field) using a clear search strategy (e.g., listing your search terms and sources) so that the process can be easily replicated.
When conducting a meta-analysis, you will begin with a systematic literature review to gather your data.
Example: In a systematic review about consumer acceptance of clean meat, researchers looked at 14 studies published in peer-reviewed journals, coming to the conclusion that consumer rejection of clean meat is driven by concerns about taste, price, and safety.
Meaning: This is a way of analyzing data collected in qualitative research. The analyst reads through the codes generated for a set of interviews, focus groups, open-ended survey questions, or other qualitative data, looking for patterns or themes in what people said. Multiple codes may be grouped together to form a theme.
Example: In Faunalytics’ poll about the public’s knowledge of the relationship of the COVID-19 pandemic to animals, we asked people to explain their understanding of where the disease came from. We analyzed coded responses to this open-ended question to find common themes in participants’ beliefs about the origins of COVID-19. In this case, the thematic analysis was computer-assisted: we used a program to automatically code key words in participants’ answers and then reviewed them to create themes.
Meaning: In many experiments the treatment is what the researchers want to test the effect of. People in the treatment condition are exposed to an experience involving that treatment of interest, while people in the control condition have a more neutral experience or nothing at all.
Example: In our experiment on the effectiveness of video outreach on pork consumption, the treatment group watched a video on factory farming while the control group didn’t.
Meaning: This is a particular statistical test that tells you whether there is a statistically significant difference between the average scores of two groups.
Example: In our study on animal advocate experiences in the U.S. and Canada, we ran t-tests to compare things like paid vs. unpaid advocates’ average levels of burnout.
Meaning: The extent to which a tool (like a survey question) accurately measures what you intend to measure.
Example: To be valid, a measure of speciesism should reflect the belief that humans are superior to other animals, rather than another animal-related attitude such as liking or attachment.
Meaning: Any characteristic of a person, place, or object that can be measured and varies between people.
Example: Variables include everything from demographic characteristics (e.g., gender, age, nationality) to attitudes (e.g., support for animal protections) to behaviors (e.g., frequency of consuming meat).
Meaning: As the term suggests, variance measures the variability of your data (similar to standard deviation — it’s actually standard deviation squared). In other words, it tells you how spread out your data is from the mean. So, the larger the variance, the more spread out your data is.
Researchers looking at whether uncertainty is reported in population estimates for endangered animals found that not many species recovery plans in the U.S. specify variance in their estimates of population size, which can set the stage for questionable recovery criteria.
In our Original Studies, you’ll mainly see variance mentioned with regard to a regression coefficient.
Meaning: While not technically a research term, this comes up a lot at Faunalytics. We use this short form to stand in for “vegan and vegetarian” when referring to both groups collectively. While vegans and vegetarians are different in many ways, they are often much more similar to each other than to the majority of a population who are meat consumers. Combining the groups sometimes gives us more confidence in the statistical inferences we make because it increases the sample size.
Example: In our Going Veg*n study, Faunalytics looks at people’s motivations and influences for switching to a vegan or vegetarian diet.
Meaning: This term is a formal statistical version of a general concept: In everyday life, you might say you give more weight to one person’s opinion than another’s. In a survey context, we will sometimes weight data from some respondents more heavily than others to make the results more representative of the population those respondents come from.
Example: In Faunalytics’ experiment to see whether messages and donor characteristics can increase donations to farmed and companion animals, people of color were under-represented and people ages 25–44 were over-represented. For this reason, responses from people in those groups were weighted a bit more heavily to make the data more representative of the general population.