My professor wants the writing to be about Bakersfield, California that is where I grew up. I used to work in agriculture in grapes and blueberries.
Here are the choice readings the three articles I choose:
Bissing-Olson, M.J., Fielding, K.S., & Iyer, A. (2016). Experiences of pride, not guilt, predict pro-environmental behavior when pro-environmental descriptive norms are more positive, Journal of Environmental Psychology, 45, 145-153.
Bratanova, B., Loughnan, S., & Gatersleben, B. (2012). The moral circle as a common motivational cause of cross-situational pro-environmentalism. European Journal of Social
Psychology, 42, 539-545.
Thompson, C. W., & Aspinall, P. A. (2011). Natural environments and their impact on activity, health, and quality of life. Applied Psychology: Health and Well-being, 3(3), 230-260.
Terms you can discuss in the paper:
Threats to internal validity include:
Sample—the people who are actually measured in a study—a big question is how well the sample represents the population (therefore how well can your findings generalize to the population?)
Representative sample—using population data, researchers select a representative slice of the population, making sure the distribution of certain variables such as income are proportionally represented (e.g., well-done polling)—almost never seen in psych studies
Random sample—given that all of the members of a population are known, select a sample so that each person has an equal chance to be selected—almost never seen in psych studies
Convenience sample—using a convenient sample of the population—such samples are seldom fully representative of the population—this is what most psych studies do (e.g., the infamous “college first-year” sample). However, the extent to which this is an issue depends largely upon whether the variables probably vary depending upon age, ethnicity, education, or gender.
Snowball sample—given a narrowly-defined population (e.g., middle-aged feminist vegans), recruitthe sample by locating a few individuals who then refer the researcher to others they know—a few psych studies do this because of the difficulty in finding such unusual people.
Meta-analysis—the sample is actually a large collection of relevant studies, with each study being a “data point” (the population is all studies fitting criteria set by the researcher, so are the criteria capturing a good sample of the “population” of studies on this topic?)
Biased sample—whether the researcher is aware of it or not, some samples are skewed in such a way that they clearly do not represent the population—e.g., historically, psych has used samples that over-represented males and white middle-class college-educated people and under-represented females, the poor, and ethnic minorities (in studies intended to represent the populations “people” or “children”)although this problem is not as glaring as it was decades ago—and a study doesn’t have to have an exactly equal number of people in each variable category, esp. if there’s some reason to think that the samples’ characteristics are not relevant to the behavior/process being studied. Also, if the population does not have an equal proportion of people across categories of interest, the sample can be representative of the population by reflecting these proportions.
Selection bias—when you are comparing groups, the groups systematically vary to begin with in some important but unintended way, e.g., the people in one group are older than the people in another—a confound–—a fairly frequent result of non-random samples
Selective drop-out—in studies occurring over time (pre-post or longitudinal), the people who drop out of the study systematically vary from those who stay in some important and unintended way
Weak face validity—when moving from the concept intended to be measured to the actual measure of the concept (operationalization), the measure or procedures/design just don’t seem to capture the concept very well to begin with (e.g., if you ask people if they consider companion animals family members and 98% say “yes” that’s nice but you don’t have a variable (it’s not varying!))
Weak construct validity—the study overall, or its measures, don’t tie in well with any particular theory—maybe you find out what happens but you have no theoretical base for understanding why, e.g.
Weak convergent validity—several measures are used that theoretically/conceptually should be closely related, but they don’t turn out that way
Weak discriminant validity—several measures are used which are not theoretically/conceptually related, yet they do not clearly show different relationships
Mono-method or mono-measure issue—only one measure or technique is used to capture the variables, so the coverage of the concepts involved is limited
Testing effects—when multiple measures are used, or measures are repeated, there may be practice effects or an influence of taking one measure on completing some later measure
Reactivity—the members of the sample react in some important and unintended way to the measures or procedures—including response sets (e.g., tendency to pick the middle of rating scales), social desirability effects (e.g., reporting “better” attitudes and behaviors than one actually has), demand characteristics (e.g., people may do something for a researcher that they would not actually consider
doing in day-to-day life)—researchers should often be trying to design the study or measures in such as way as to keep the participants “blind” to the aims of the study or to test, within the study, for things such as social desirability effects.
Low reliability—the measures used are not very good for measuring what they’re intended to measure
including, e.g., ceiling or floor effects—there is not much variation among the people in the sample in their responses—as a group they tend to max out or bottom out in the possible scores—or the measure has low internal consistency or low repeatability
REVIEW OF OTHER CONCEPTS
Variables—in psychology, variables are characteristics and behaviors (in the broadest sense) that vary across people—studies ask how variables co-vary, i.e., how well knowing a person’s “value” for one variable helps predict the person’s “value” on a second variable, or how variables vary across groups
Defining variables—researchers need to state what exactly the variable is conceptually and then operationalize it validly and reliably—operationalizing is a crucial, central part of research—you can never capture the entire concept withyour measures and procedures, but you can do it well or poorly—in a study, the researcher moves from a set of conceptual measures to clearly operationalized variables (stating how variables are constructed–from what measure, what items, how it is calculated (a sum, an average), and its possible range and scaling)
Operationalized variables take the form of nominal, ordinal, or interval/ratio scales in quantitative studies
Nominal–2+ categories–e.g., ethnicity categories, age as categories, using a single item with yes/no answers as a variable
Univariate description=frequency (# and % of people in each category)
–you cannot treat nominal variables as interval/ratio–they do not have means, SDs, etc.
When a number system is used for a nominal variable this is arbitrary and not to be used mathematically (e.g., Democrat category =1 and Republican category = 2 for scoring purposes—you don’t take averages or do other math computations on these numbers.
Ordinal–values that are ranks–in reality, most Likert scales or similar measurements are ordinal but in practice we usually treat them as interval or ratio
Interval or ratio–true numbers, with ratio scales have a meaningful zero point–the values mean something mathematically (2 is twice the value of 4, e.g.)–you can add, subtract, multiply and divide them meaningfully, e.g., age in years, total number of yes answers on a 20-item survey, rating for fun experienced in a game on a numbered scale
Although Likert ratings are actually ordinal, we usually treat them as interval/ratio
Univariate description=mean and SD (pay attention to SDs!—are participants clumped together or spread out? and pay attention to the mean and what it says about the overall sample—did the sample have a mean that’s high? middle? low? for what it could be?)
Operationalized variables in qualitative studies often take the form of “themes” (software exists to help researchers find the themes)
Independent variable—the predictor variable, the one the researcher uses to try to predict how people vary in the dependent variable, i.e., how well does knowing a person’s value on the independent variable predict the dependent variable value?
Subject variables—some variable values are inherent in a person and cannot be “tweaked” by researchers in a study (e.g., age, sex, ethnicity, religion, political party…)
Dependent variable—the predicted variable, the one the researcher wants to discover, given the value of the independent variable
Hypothesis—a conceptual (or general) hypothesis is a fairly informal statement about how concepts are related (e.g., “Because of gender socialization, women are likely to be empathic”) while studies actually test a formal hypothesis that incorporates operationalized variables (e.g., “There will be a relationship between sex and score on the XYZ Empathy Measure such that females score higher than males”).
Statistics—simple statistics (one IV and one DV), to a great extent, follow from the scaling of the variables
2-category nominalint/ratioindepsample t-test
3+-category nominalint/ratio1-way ANOVA
nominal (2-3 cats.)nominal (2-3 cats.)chi square (need cross-tab table
to interpret findings)
Other statistics—Factor analysis—correlates each item in a measure with every other item in order to find
“clumps” or factors of items that are strongly related to each other—often used to test or build theories and to
refine measures (e.g., in an intelligence test, how many factors are there and what are they?)—researchers often
argue about the content, number, and interpretation of factors in a measure/concept—
MANOVA (Multiple Analysis of Variance)—one IV and multiple DVs, testing how well the DVs hang together as a “set”—
Multiple regression—several IVs used to test the best predictive combination for one DV—
Discriminant analysis—tests whether groups of people can be reliably distinguished from each other using certain variables or measures—for instance, can you use peoples’ political beliefs to predict with a high degree of certainty whether or not they recycle?—the analysis “backtracks” by forming recycling versus not recycling groups and then testing which IVs best predict (discriminate) the group membership
Significance—quantitative findings indicate significance in a way that tells you how likely it is that the findings are
due to random chance and errors—psych accepts assignificant findings with fewer than one in 20 chances of
being due to chance (p