Chapter 1: Why do we use Statistics?
Steps in the Research Process
1. Choosing a research question 2. Conducting a literature review 3. Developing a hypothesis 4. Designing the study 5. Conducting the study 6. Analyzing the data 7. Reporting the results
1.7 Consider the differences between experiments and correlational studies as you read the research questions below. For each question, which type of study (experimental or correlational) would be better for answering the question? Explain your answers. 1. Do energy drinks help you focus more while studying? 2. Do anxious people tend to sleep less? 3. Does eating red meat give you cancer? 4. Who earns a higher starting salary in their first job, people with higher college grade point averages (GPAs) or people with higher aptitude test scores (ACT, SAT, etc.)?
1. Experimental. You could set up a study in which energy drinks are the independent variable. 2. Experimental. The level of anxiety would be the independent variable in a two-group study. Anxiety could be determined through self-report or some other way. 3. Correlational. Because you can't realistically impose the consumption of red meat over a lifetime, a comparison of dietary data and medical files is more appropriate. 4. Correlational. You're looking for a relationship between two types of scores and salary, not a cause.
1.9 Suppose you come across each of the surveys described below. Which issue seems to be the bigger problem, validity or reliability? 1. You see a survey in a magazine about the quality of your relationship with your significant other. The items on the survey ask questions about your favorite color, favorite food, and favorite type of music. 2. You complete a survey as part of a research study on eating behaviors of college students. The items ask how much you like to eat different types of foods. You complete the survey once while you are hungry and then again one year later after you have eaten a large meal. Overall, your ratings are lower the second time you take the survey.
1. Validity 2. Reliability
What are the methods we use to collect data?
A brief description of some of the methods used in collecting data was provided in this chapter with an emphasis on the differences between experiments that provide causal information and correlational studies that provide information about relationships between different types of measures. The statistics we use to test our predictions about the data will depend on the methods used to collect those data and the scale of those measures.
experimental design
A design in which researchers manipulate an independent variable and measure a dependent variable to determine a cause-and-effect relationship
positive relationship
A relationship in which the values of one variable increase (or decrease) as the values of another variable increase (or decrease) (both go up together and both go down together)
Reliability
Ability of a test to yield very similar scores for the same individual over repeated testings; consistency of measurement
How do the methods used to collect data affect the statistics we use?
As already mentioned, experiments and correlational studies use different inferential statistics because data are collected to answer different kinds of research questions in these designs. In addition, the observation techniques can vary across these designs, which require different types of statistics to better understand them. For example, in survey studies, there are statistics to help us examine the validity and reliability of the survey. The rest of this text will discuss this question in much more detail.
How do descriptive and inferential statistics differ?
Descriptive statistics help us summarize a set of data. They include graphs and tables of the data, calculated values that represent typical scores, and values that represent the difference between the scores. Inferential statistics help us test hypotheses made about the data. They use the descriptive statistics to determine the likelihood of obtaining our data when a hypothesis about the data is true.
identify the independent and dependent variables: Your statistics instructor has recruited students to be in a study in his lab. You sign up for the study, and when you participate, this is what you are asked to do: You are asked to complete two blocks of trials where you have to decide if a string of letters that appears on the screen is a real word or not as quickly as you can. During one block of trials, you focus entirely on this task. In the other block of trials, you are asked to also hold a short list of words in memory until the end of the block, when you have to recall them. You are told that the purpose of the study is to examine the effect of the memory task on your ability to decide if the strings of letters are words. identify whether it is an experiment or a quasi-experiment. identify whether the independent variable was manipulated between subjects or within subjects.
Independent: List of words meant to be recalled later Dependent: Ability to decide if strings of letters are words Experiment Within Subjects
identify the independent and dependent variables: You want to know if having a cell phone out while you study is a distraction, so you conduct a short study to figure this out. You observe your friends while they study for their coursework. You record whether each one has a cell phone out while they study and place them in either the "cell phone" group or "no cell phone" group, based on what you observe. Then you record how many minutes out of an hour of studying they appear to be on task. You compare the two groups of people to see if they differ in time on task. identify whether it is an experiment or a quasi-experiment. identify whether the independent variable was manipulated between subjects or within subjects.
Independent: Presence of a cell phone Dependent: Time studying on task Quasi-experiment Between Subjects
Consider the description of the Inzlicht et al. (2006) study presented in this section. How could these researchers have examined a causal relationship between stigma and self-regulation? Why do you think they chose to conduct a correlational study instead?
Researchers could have designed a study in which they regulated distracting events (as the independent variable) in order to measure the students' self-regulation. The stigma sensitivity of the African American students would have been recorded before the administration of the stimuli. But an 437 experiment of that kind would be unethical. A correlational study is more appropriate.
Why do we use statistics to analyze data?
Statistics help us summarize a set of data and test hypotheses about behavior. They are important tools in understanding data from research studies in which we learn new knowledge about behavior.
Explain how an unreliable measure of behavior can lower the internal validity of a research study.
Take a study that seeks to determine how much water college students consume on Tuesdays. The poll is given four times a year on a Tuesday. The measure is unreliable because, if the students live in a location with drastically different seasons, the amount of water consumed may vary considerably. That is, more water is likely to be consumed on the Tuesday in summer than the Tuesday in winter. The internal validity, then, is low. The day of the week may not contribute to the amount of water consumed. Other factors may be at work.
Consider some statistics you have encountered recently in your daily life. In what way(s) have these statistics influenced your thinking about an issue?
The New York Times is running a graphic on its website that gives the probability of who will win the November presidential election. It changes daily (or more often). Yesterday, it gave an 88% chance to Hillary Clinton and 12% chance to Donald Trump. Such a graphic makes me think the election is essentially wrapped up.
Validity
The ability of a test to measure what it is intended to measure
independent variable
The experimental factor that is manipulated; the variable whose effect is being studied.
dependent variable
The measurable effect, outcome, or response in which the research is interested.
Why are there so many different kinds of statistical tests?
There are many different ways to observe behavior, so many statistics have been developed to help researchers understand the observations that they have used. In addition, different statistics are helpful for the types of research designs described in this chapter. For example, experiments and correlational studies rely on different types of inferential statistics to answer the research questions asked in each of these designs.
True/False: correlational studies do not provide direct causal information about the relationship
True
True/False? Experiments will help us determine if something causes a behavior, but correlational studies can only help us determine if there is some kind of relationship (which might not be a causal relationship) between two measures.
True
True/False? independent variables are specific to experiments and quasi-experiments.
True
Review the statistics presented from the New York Times on the decrease of heart attack rates. What additional information do you think you would need to apply this rate reduction statistic to an individual's current heart attack probability?
We would need more detail about the population polled. What were the ages of the participants? The exercise habits? Diets?
Review the description of the study about teen smoking rates. Based on the statistics presented, would you recommend a national increase in the age to buy cigarettes to 21 years? Why or why not?
Yes, I would be in favor of a nationwide increase. Based on the statistics, the implementation of higher legal smoking age cut down on purchases by people under 18 years old. The health benefits would be worth it.
internal consistency
a form of reliability that tests relationships between scores on different items
inter-rater reliability
a measure of the degree to which different observers measure behaviors in similar ways
negative relationship
a relationship between variables characterized by an increase in one variable that occurs with a decrease in the other variable (as one goes up, the other goes down).
correlational study
a research project designed to discover the degree to which two variables are related to each other
experiment
a type of research design that involves the comparison of behavior observed in different situations
quasi-experiment
a type of research design that involves the comparison of behavior observed in different situations, but where subjects are not randomly assigned to the different situations An experiment in which investigators make use of control and experimental groups that already exist in the world at large. Also called a mixed design.
Variables
attributes that vary across individuals and situations, such as age, sex, and popularity
between-subjects variable
changing situations across different groups of subjects in a research study each participant experiences only one level of the independent variable
Within-subjects variable
changing situations within a single group of subjects in a research study such that each subject experiences of all the different situations being compared each participant experiences all levels of the variable
internal validity
extent to which we can draw cause-and-effect inferences from a study; the degree to which a research study provides causal information about behavior
Frequency
how often a response or score occurs within a data set
test-retest reliability
indicates that the scores on a survey will be similar when participants complete the survey more than once under similar circumstances
descriptive statistics
statistics that help researchers summarize or describe data
Mean
the average score for a set of data
construct validity
the degree to which a measure is an accurate measure of the behavior of interest
external validity
the extent to which the results of a study can be generalized to other situations and to other people; the degree to which the results of a study apply to individuals and realistic behaviors outside the study
Variability
the spread of scores in a distribution
no relationship
the two dependent variables do not consistently change together.