Please enable JavaScript.
Coggle requires JavaScript to display documents.
Concept Map - Cate and Luna (Vocabulary of experiments (Factors - These…
Concept Map - Cate and Luna
Sampling Error
- This type of error can be seen when a statistical characteristic about a population is made from a sample, not the group as a whole.
Convenience Sampling
- This is taking the members of the population the most convenient for the researcher. With this sample, unrepresentative data is often found due to the sample being chosen solely on their reach. An example of this is a surveyor goes to the bar and asks if there should be a legal amount of alcohol each individual should be able to have. Due to the surveyor choosing convenience over wanting to actually collect a fair sample, they will most likely get a biased response.
Voluntary Response Sampling
- Even though this sampling is voluntary, there is still a high basis surrounding their answers. Due to the individuals being inclined to have strong opinions, their opinion would be biased. An example where this can be seen is when people call into a specific radio station and may have a strong opinion about the poll they may be discussing.
Undercoverage
- This occurs when a surveyor leaves out a certain group in a location. Adding this factor means that their needs to be a specific sampling frame with a list of every individual in that population. Many samples have undercoverage due to the difficulty of getting an entire population to take a survey. A classic example of this was when the literary digest voter survey predicted that Alfred Landon would beat Franklin Roosevelt in the 1936 election.
Non-sampling Error
- This type of error can be seen when there are errors made due to how the survey is conducted. There may be a really good random sample, but the survey will still conatin some sort of bias. An example of this would be for if an individual went to a state where there are mostly republicans. This error would cause for there to be a bias towards the views of a republican.
Wording Bias
- As seen in many examples through the media, this is when you influence the way an individual thinks/reacts about a specific topic. An example of wording bias is, Dogs carry harmful diseases, so they shouldn’t be allowed near a sports field where there are people of all ages.
Response Bias
- This is an individual tendency to answer questions on a survey untruthfully or in a misleading manner. For example, people know that they should vote, so individuals that didn’t vote in the last election could go and tell an interviewer that they did. This is an example where individuals may feel socially pressured to respond in a certain way towards a controversial topic.
Non-response
- When a large number of people say that they took a survey when in reality they didn’t, so it disrupts the data and gives false information. An example as to why this might happen is that the survey did not reach all the individuals it was intended for.
Bias
- It is a mistake that the surveyor makes that under predicts a population. An example is of an individual went to a gun range and asks about whether America should have stronger gun laws, there would be a bias for not any gun control. Due to this sample being convenient, there will not be a good predictor of the response. Sources of when a bias occurs is the sampling and non-sampling erros that can be made.
Sources
When an individual uses specific words to influence their decision about the topic that the surveyor is asking.
If an individual answers questions on a survey untruthfully or in a misleading manner.
When we make a mistake on how we get the information. Sampling and nonsampling errors bias is always associated with a direction.
If an individual said that they took the survey and gave their response when in reality they didn’t.
When an observer doesn’t include either the response or explanatory variables, but can influence the the interpretation of relationships between the variables
When there is another variable unaccounted for, which can suggest that there is a correlation when there isn’t.
Direction
- There are many occasions where bias can be directed. For example, if there was a scientist that was testing the effects of a certain drug, they could take the results that were positive about the drug and represent it. This was would be an inaccurate skew and make the nformation about the prosuct appear to be an effective drug.
Good Sampling techniques
SRS
- The basis for a random sampling is often referred to as a simple random sample (SRS). This is an opportunity for there not to be any bias towards the information. An advantage of using SRS is that everyone in the sample has the same probability to be chosen. A disadvantage would be the time needed to obtain the information and the bias that could occur when a sample set is not large enough to to represent the population being surveyed. An example of SRS is the asking a quation to people that are attending a Patriots game at Gillette Stadium. Due to there being a large group of people, there needs to be an SRS made to get a nonbiased response to the question.
Stratified SRS
- This occurs when the surveyor separates the individuals into smaller groups, also known as strata. These smaller groups are often similar in one way or another because they may affect the responses to a survey. An advantage to this type is that there are individuals in a smaller group that have common interests, so their responses will be similar. A disadvantage to this is that there most likely be one perspective/bias from this smaller group.
Cluster Sampling
- To choose a cluster sample, one must divide the population into groups, or clusters. Randomly select some of these clusters. All the individuals in the chosen clusters are included in the sample. An advantage of this type of sampling is that you can find clusters in a geographic area to sample if you don’t have access to the population as a whole. A disadvantage of this type of sampling is that it’s not as precise as other methods, for example an SRS.
Design Survey example
An example of a simple random sample would be the names of 15 teachers being chosen out of a hat from a school of 150 teachers. In this case, the population is all 150 teachers, and the sample is random because each teacher has an equal chance of being chosen. The use for this is that a surveyor could find unbiased information from 15 teachers since everyone would have the equal probability of being chosen. The numbrs that I would be looking at are teachers 20-35.
(Chart)
https://docs.google.com/document/d/1KCK53cBr806Ju6MNMNTys6kihXxrDuLtaWd2ExBNXW8/edit?usp=sharing
Observational Studies vs. Experimental Studies:
An observational study observes individuals and measures the variables of interest, while the experimental studies are when an individual artificially manipulates a specific subject.
Observational Studies: An observational study observes individuals and measures variables of interest. A difference in this type of study is that the observer doesn't attempt to influence the responses. An example of an observational study would be giving out a survey about people’s opinions on genetically modified produce.
Experimental Studies: An experimental study occurs when an individual artificially tries to manipulate a specific subject. An example of an experimental study would be giving a group of people a new medicine to see if it has more or less beneficial effects.
How can a lurking variable lead to confounding?
A lurking variable occurs when the observer doesn’t include either the response or explanatory variables, but they can influence the interpretation of relationships between variables. A negative factor that can be associated with this type of study is they can falsely identify strong relationships between variables or hide true relationships between variables. Confounding occurs when two variables are associated in such a way that their effects on a response variable cannot be distinguished from each other. Some people call a lurking variable that results in confounding, a confounding variable. An example is if the effects of two variables on a response variable cannot be separated from each other, such as if age and hormones in an observational study influenced heart attacks.
Vocabulary of experiments
Factors
- These are the controlled independent variables that are used in an experiment. They are often referred to as being a general category of treatments.
Treatments
- A specific condition applied to the individuals in an experiment is called a treatment. If an experiment has several explanatory variables, a treatment is a combination of the specific values of these variables.
Experimental Units (Subjects)
- The experimental units are the smallest group of individuals to which treatments are applied. The units are often called subjects when they are human beings.
Response Variables
- This variable measures an outcome of a study.
Important defintions
Placebo effect
- This occurs when there is a drug/treatment that has no benefit (no active ingredients in it) and cannot be attributed to the placebo had by itself. Therefore proving that their response to this “fake” drug must be through the patient's beliefs about the effects that the drug/treatment was supposed to help.
Blinding
- This term refers to when an experimenter intentionally keeps the patients in the dark about knowing if they are receiving the placebo drug/treatment or not.
Three principles of experiment design
Randomization
- This helps to ensure that the estimate that is made about the individuals being tested is statistically valid. The principle is important because it minimizes the bias in responses to an experiment.
Repetition
- This part of the experimental design helps to provide an estimate of the experimental errors that may occur in the data. As seen in every experiment, the repetition of an experiment is important because it allows for the data collected to be more accurate. The more an experiment is repeated and similar results happen, the more accurate the conclusions are.
Control
- This part of the experiment design is used to help reduce the probability of an error by ensuring that the experiment is efficient. It is important to the design because it allows the experimenter to have an increase in reliability for the result of the experiment.
Our example
for a completely randomized design for an experiment tests whether certain applications on a laptop influences the battery life of a laptop. Our original sample size was 50 laptops, which changed after using the random number chart.
We first used a random number chart to identify the numbers that were between 20-50. After looking at row 1 of the chart, the first two numbers that could be used for the experiment were 48 and 30. After completing the information we needed to collect from the chart, we implemented the design, which was to start by labeling each laptop with a distinct number from 1 to 30. We would then use SRS to randomly assign each laptop to a group of 10. We would then draw 10 numbers. The corresponding laptops would be used only with google chrome. The next 10 laptops would only be used with Safari. The remaining 10 laptops could be used without either of those applications. Battery lives would have to be recorded by each laptop and at the end of one week compared, to see how much battery life was being used by the laptops in the three groups.
A case where blinding could be used in this experiment was the application that was put on the laptop. Through this, the individuals would not be able to tell of their battery life was affected by an application used in the experiment. A placebo that could also be used in the experiment is the application that the experimenter put on the computer. If an individual has already used the application then there could be a bias towards the software that is said to be on the laptop.
(Chart)
https://docs.google.com/document/d/1EzKy9JJ97hLYYp_zyx0pNN9enNdKe-1ZqQ_eHLzqmaw/edit?usp=sharing
