#statistics studyblr

9.1.2021

Research Methods

In some ways, Psychology is a science. It is difficult to define it as a science since some approaches cannot depend on the scientific method to perform experiments due to the unpredictable nature of the human psyche. While in Biology, it is easy to isolate two variables to study, however, in Psychology some complications such as hindsight bias can appear. Hindsight bias is the tendency to assume that you know the results all along. 

There are several ways psychologists can use psychology to change the world. Some conduct research to solve practical problems. This is called applied research. Others may conduct basic research, which is when psychologists conduct experiments not intended to have immediate real world effects.

Terminology

Most psychological research is guided by a hypothesis. It expresses a relationship between two variables, which are things that can vary among participants in an experiment. Skin colour, age and hair texture are all variables. The dependent variable is the variable who depends on the independent variable (hence the name.) The independent variable is typically the variable being altered in some way, which impacts the dependent variable which is typically the variable being studied. For example, if you wanted to look and see if the heat of the environment impacts the structure of a protein, the dependent variable is the protein structure, and the independent variable is the temperature. A theory attempts to explain certain phenomena, and gives scientists the ability to generate testable hypotheses. Psychologists also have to provide an operational definition of the variable that they will study. This means that they will explain how they measure that variable. 

Good research is valid and reliable. To be reliable means that the research can be replicated, and to be valid means it measures what the researcher wants it to measure. For example, if someone made Albert Einstein take an IQ test which rated his intelligence from 1 to 100 to find out if he was above or below average intelligence if he were to get a 9 consistently, the test is clearly reliable but invalid. If Albert Einstein were to take this test and get a 100 then a 9, then this test would be unreliable and invalid.

Before a hypothesis is investigated, who or what is being studied needs to be decided. Participants are the individuals who the research is conducted on, and the process that participants are chosen is called sampling. The goal of sampling is to pick a representative group to allow for generalisations to be made. Random selection is ideal as every member of the community has an equal chance to be chosen. (In a way that does not allow for bias so that the selection is truly random). Stratified Sampling is another process that makes sure a sample represents a population. For example, if in my study I am including two religious groups: Religion A and Religion B, and I feel that religion may have some bearing on my research, I could divide the group of possible participants into two groups and then randomly choose people, thus ensuring representation from two groups which may affect each other.

Experimental Method

There are two kinds of experiments: laboratory experiments and field experiments. Laboratory Experiments are done in a lab (hence the name) which is an extremely controlled environment. Field Experiments are performed in the real world. Laboratory Experiments allow for more control, while Field Experiments allow for more realism. Experiments are the preferred method of research among psychologists because it is the best way to show a casual relationship. It allows manipulation of an independent variable and to control for confounding variables which is any difference between the experimental and control conditions except for the indpendent variable which can cause changes in the dependent variable. If I want to prove that cheese causes nightmares, it’s best to perform an experiment where I make sure that it is the cheese leading to the nightmares, and not other factors such as my subjects being sleep deprived, or being forcefed ounces of cheese which would inevitably cause nightmares.

Assignment is the process by which participants are put into a group (either experimental or control.) Random assignment means that each participant has an equal chance of being placed into any group. Random assignment reduces the effect of participant-relevant confounding variables. If in my cheese experiment, I gave my participants the option to choose whether they would like to be the control or experimental group, people who don’t like cheese would be more likely to avoid the experimental group, meaning the people eating massive amounts of cheese likely wouldn’t be scarred by eating ounces of it. Group matching can be used to make sure experimental and control groups are equivelant on some criterion. For example, if I want to make sure I have equal amounts of men and women in my groups, I could split my group into men and women, and then assign them to control or experiment. 

Situation-relevant confounding variables can also prove to be an issue. The situations that the different groups are put into must be equivelant except for the independent variable. These can include the weather, the ambiance, or the lighting. Experimenter Bias is a specific type of situation-relevant confounding variable. It is the tendency for the experimenter to treat members of the control or experimental group differently to increase their chances of confirming their hypothesis. (It isn’t a conscious action). Double-blind procedures can eliminate this bias. Normally, the way these work is that a researcher has someone who doesn’t know who is control and who is experimental interact with the subjects. Single-Blind procedures are when only the subjects don’t know which group they are assigned to. This minimises demand characteristics and types of response and participant bias. Demand characteristics are cues about what the study’s purpose is. Participants may try to use these to figure out how they are supposed to behave. One type of response bias is social desirability which is the tendency to give answers which will reflect best on the person’s character.

Control groups give something to compare the results of the experimental group to, thus concluding whether there is a relationship between the two variables. In fact, simply selecting a group of people to experiment on can affect the performance of the group regardless of what is done. This is what is called the Hawthorne Effect.

Another method of control is called the placebo method. This is typically used in drug experiments. The control group are given a sugar pill- that looks identical to the functioning pill. This allows psychologists to seperate the physiological effects of the pill to the psychological effects known as the placebo effect.

In some cases, participants can be used as their own control group- this is known as counterbalancing. This can be flawed as it can cause order effects. For example, if a psychologist was seeing how cheese affects peoples math skills, they can give the participants a math test, feed them cheese, and then give them a second math test, however they may do better on the second test simply because they’re doing it over again. Counterbalancing helps deal with this problem as half of the group gets the cheese first, and the other half doesn’t.

Correlational Method

Correlation expressed a relationship between two variables. It does not imply causation. 

Click here: https://twentytwowords.com/funny-graphs-show-correlation-between-completely-unrelated-stats-9-pictures/ to see some very bizarre examples of this very important rule when using the correlational method. A positive correlation between two things means that they have a direct relationship (If I do more of Action A, I will get more of Object B) if they have a negative correlation, it implies an indirect relationship (If I do more of A, I will get less of B). 

Some hypotheses are impossible to make into experiments, which is when the correlational method can come in handy. These are situations where the independent variable is impossible to control; for example if I want to behave differences in men and women’s behaviour, I can’t control the independent variable because it will be predetermined, meaning I can’t guarantee differences in behaviour can be ascribed to gender differences or other phenomena I can’t account for with random assignment. If I try to control all other parts of the research in this case, it is considered an ex post facto study. 

Another popular use of the correlational method is the survey method. For example, If I want to see if wearing socks leads to depression, I can ask people to fill out a survey detailing if they have depression and how often they wear socks. This method, like any correlational method cannot identify a cause-effect relationship, only establish a relationship. (I don’t know how many ways I can say this. Stop using correlational studies to prove your bogus cause-effect relationship hypotheses. Vaccines don’t cause autism.) There is no independent or dependent variable in a survey. One major flaw with the survey method is that there is no way to control participant-relevant confounding variables. They can be massively convenient, as a psychologist can have several people mail back responses, however, a random sample is difficult to obtain because few people will send it back (it has a low response rate)

Naturalistic Observation

Naturalistic observation is unobtrusive. It provides no control, however gives the researcher a detailed look into the behaviour of a person. One example is the case study method where researchers spend years studying one, or a small group of people, giving intimate details into how their lives are and how certain events may lead to a certain behaviour they want to observe. These types of studies cannot be generalised due to their small sample size.

Statistics

Now that you have all your data, you need to be able to interpret it, otherwise; what’s the point of having all that data? Descriptive statistics is a creatively named part of statistics where stats are described. You can summarize your data with a frequency distribution. Take an example where you were observing what colour everyone in the office wore one day. A frequency distribution would tell you how many people wore red, how many wore black etc etc. You can graph this data and turn it into line graphs called frequency polygons or bar graphs called histograms. Frequency always goes on the y axis.

Another way to summarize a set of data is by looking at the central tendency. We can do this by looking at the mean (adding all the numbers together and dividing by the number of scores) the median (write all the numbers in ascending orders, then find the middle number, or if there is an even amount of numbers, the average of the two middle numbers) and the mode (the most frequently appearing number). Some sets of data have more than one mode. A bimodal has 2 scores appear equally frequently and more frequently than the other scores.

The mean, while the most commonly used can be affected by outliers. Outliers are scores that are way outside of the norm. For example, take GPA, which is the average of all of your grades. Say for 4 years you were a straight A student in all 6 of your classes, except one year, you flub and get an F in math. Your GPA would be a 3.33, implying that you are a B average student. Your GPA was negatively skewed by that one bad year in math, meaning that the skew was caused by an extremely low score. If you were a straight D student, but one year managed to get an A in both Math and English, your GPA would be a 2.00, implying you are a C average student. This data is said to be positively skewed, as an abnormally high score caused the skew. In unscrewed data, the median and mean are exactly the same. With negatively skewed data, on a graph, the mean shift left, while with positively skewed data, the mean shifts right. 

Measures of variability can also describe the data you have collected. These include range, variance, and standard deviation. These depict the diversity of the data. Standard deviation is the square root of the variance. They relate the average distance of the scores from the mean. A higher variance and SD implies more spread out data. The range is the difference between the highest and lowest score. If you want to compare scores from different distributions, you can convert the scores into z scores. These measure the distance of a score from the mean in units of standard deviation. Say a student got a 72 on a test where the average score was an 80 with a SD of 8. Her z score would be -1. If someone else got 84 on the same test, her z score would be +0.5. The normal curve is a bell shaped curve, where the area under the curve lying between two z scores has been predetermined. Z scores calculate the distance of a score from the mean, while percentiles indicate the distance of a score from 0. Someone who scores a 1250 on their SAT is around the 81st percentile, meaning they scores Better than 81% of test takers. Someone in the 50th percentile has a z score of 0. Someone with a z score of 98 has a z score of 2. See the pattern?

Correlations

Correlations observe the relationships between two variables. They can either be strong or weak; this can be calculated with a correlation coefficient. These can range from -1 to +1, with -1 being a strong negative correlation and +1 being a strong positive correlation. 0 means that there is no correlation. Correlation can be graphed with a scatter plot. If there is a strong correlation, the dots on the scatter plot will come close to a straight line. The line of best fit (also known as the regression line) is a line drawn through the plot that minimizes the distance of all plots from the line.

Inferential Statistics

Inferential Statistics determine whether findings can be generalised to a larger population. It is impossible to guarantee that a sample is representative of the entire population. The extent of the difference between the sample and population is called sampling error. There are many tests that determine if findings can be applied to a larger sample of people such as chi square, t-tests, and ANOVAs. These tests yield the p value which observes the probability that the results were due to random chance. A p value of .05 is the cut off for statistically significant results. A p value of .05 means that there is a 5% chance that the results were due to random chance.

APA Ethical Guidelines

Through its history, Psychology has lead to many horrific experiments which have caused significant harm to its patients. One example is the Little Albert experiment, done by John B Watson. He clanged an iron pole every time the little boy reached out to touch a white rat. Even without the sound, the boy was afraid of the rat showing emotions could be conditioned. While this research provided important information for scientists, had Albert not died at 6, could have had serious implications for him in his life outside of the experiment. The APA now has a series of guidelines to assure research does not cause serious harm. Before research can begin, the study must be presented to the Institutional Review Board (IRB) which reviews proposals for ethical violations and procedural errors. The regulations vary based on whether the subject is a human or an animal.

For animal research, the study must meet these requirements:

- They must have a clear scientific purpose

- The research must answer a specific and important scientific question

- Animals chosen must be best suited to answer said question

- The animals must be cared for and housed humanely

- The animal subjects must be acquired legally and purchased from accredited companies. Wild animals must be caught humanely

- Experimental design must employ the least amount of suffering possible. 

Research involving human subjects must meet these requirements: 

- No coercion. Participation must be voluntary

-Informed Consent: Participants must know that they are involved in research. If deception is necessary, the research they consent to should be similar enough to the research they participate in.

-Anonymity/Confidentiality: Participants have anonymity when no data is collected that can allow a persons response to be matched with their name. If anonymity is not possible, confidentiality is necessary, where the researcher doesn’t identify the source of their data.

-Risk: Participants can’t be placed at any significant mental or physical risk.

11.21.2021

12.9.2021

8.29.2021