Experimental Design Flashcards
How are study types grouped as?
- Observational
- Experimental
What is observational study?
Observations that are easily assessed/examined without interventions
What is an experimental study?
Where the hypothesis is tested using an intervention
How are experimental studies carried out?
By changing the variables of an experiment to determine whether a change in one affects the other
What is the meaning of intervention?
Attempts made to alter the outcome of the experiment- controlling the independent variable
What is an advantage of experimental studies?
All variables other than the desired independent variable are controlled, so you can see what effects the independent variable is having
What is peer-review?
When a scientist present their data and the proof and other scientists critically analyse it
What is the problem with information being given on the news/TV?
- Fake experts give out the information
- Results can be exaggerated
What is a disadvantage of observational studies?
It indicates that there is a causal effect even if there isn’t
What bad study design methods are used by pharmaceutical companies to show that their new drug works better than any others? Why is this done?
- Compare effects of drug against a placebo, which may indicate that drug works but doesn’t indicate if better than any other available drug
- Compare drug to best available drug but use best available drug at such a high dose that it has more side effects- making the new drug look better
What is the placebo effect?
An outcome which has been caused by having certain beliefs or assumptions
What is publication bias?
When only parts of data collected in research is shown and available to make an informed analysis of the research conducted
What is a variable?
Anything relevant to an experiment which can be measured
What comes under a variable?
- Evaluating what we need to measure (dependent variable)
- The independent variable that we believe is going to change the result we see
What is the independent variable sometimes called?
The explanatory variable
What is the independent variable?
The variable that causes a particular outcome
What is the dependent variable also known as?
The response variable
What is the dependent variable?
The variable influenced by the independent variable
When, in an experiment would you not want to/can’t control every variable but the independent and depend variables?
- May be doing an observational study
- Conducting an experimental study which needs to mimic real scenarios, where the dependent variable is affected by many different factors
What is a confounding variable?
A variable, aside from the independent variable, which also influences the dependent variable
What is the disadvantage of confounding variables in an experiment?
- It makes any results obtained more difficult to decipher
- It is difficult to pick out all he confounding variables
What can be done when all the confounding variables are identified?
- Limit/cancel affects of the confounding variables so it is more clear to see the effect of the independent variable
- Help identify the drawbacks of the study due to the confounding variables
- Further experiments can be made to accommodate for these drawbacks and strengthen the validity of the evidence provided
Why is there always a chance of error in statistics?
Statistical tests cannot completely deem a hypothesis to be true or false, it only states the chances of it happening by chance
What is error?
The difference between the data for a population and it’s sample/experiment
Why is it good to decrease the degree of error?
So that the results produced are reliable
What are the main types of error that we can control?
- Sampling error
- Bias
Where does error come about?
When we can’t measure the entire population so we take the results of a sample to represent the population
What is sampling error?
Error caused by the chances that the sample doesn’t represent the population entirely
Why are we able to estimate the sampling error?
- The sampling error is normally distributed
- Statistical tests account for take the normal distribution of sampling error
What ways are used to reduce sampling error?
- Replication
- Balance
- Blocking
How does replication control sampling error?
More replication means that there is a large data range, which increases the how accurate and valid the results may be
What are the different types of replicates?
- Technical replicate
- Biological replicate
What are technical replicates?
Repeating the same conditions of an experiment without changing anything
What are biological replicates?
The number of different individual cells/organisms being tested on
Name a difference between technical and biological replicates:
Technical replicas can happen at the same time, biological ones cannot
How does balance reduce sampling error?
- By comparing similar sized groups
- Groups of different sizes are more likely to not have equal variance
Which statistical tests assume there s equal variance between groups?
- t-test
- ANOVA
What is the effect of having an experiment with different sized samples?
- The t-test and ANOVA test would not be able to be performed
- Reduces statistical power
What function can be used for experiments with uneven sample sizes? Why does this work?
- t.test() function
- This function uses the Welch’s t-test which alters the degrees of freedom when variance between groups is not equal
What is needed to work out statistical power?
- Sample size
- Effect size (Cohen’s d)
- Significance level
How does different sized samples decrease statistical power?
- Standard deviation and variance are tightly linked
- Standard deviation is used to work out effect size
- This causes a larger combined standard deviation
- This decreases statistical power
What affect does decreased statistical power have on the probability of false negatives and true positives?
- Increase
- β= probability of getting false negative
- Power = 1-β
- If power decreases, β increases
- This decreases the chances of finding true positives
How is blocking?
Collecting experimental units that have similar properties, where treatments are randomly allocated to each experimental unit in each block
What is the advantage of blocking?
It removes the chances of an anomalous result happening due to an outside factor
What is bias?
Error caused by systematic differences in our estimation of the sample data to that of the population
At what stage of research can bias occur?
- Study design
- Data collection
- Data analysis
- Publication
How can bias occur in the study design part of an experiment?
By not randomly picking individuals/objects/
How can bias occur in the data collection part of an experiment?
Using faulty equipment
How can bias occur in the data analysis part of an experiment?
?????????
How can bias occur in the publication part of an experiment?
Results are only released if they align with hypothesis
What methods are used to control bias?
- Simultaneous control groups
- Blinding
- Randomisation
What is a control group?
A group of individuals that are not treated with the independent variable and are compared to groups that are
What are the types of control in an experiment?
- Untreated control
- Vehicle control
- Best available therapy control
- Positive control
- Negative control
What is a vehicle control? Why is it done?
- A control which is treated the same as the experimental groups but without the main independent variable
- This is to see if something else in the treatment/experiment is causing the effects observed
What is best-available therapy?
When an already existing substance /drug is compared to a sample of a new one to see which works better
What is a positive control?
A control that is know to produce the desired effects/expected outcomes
What is a negative control?
A control known not to give the desired outcome/expected effect
Why are controls described as simultaneous?
They must be carried out at the same time as experimental groups
What is blinding and when is it used?
This is typically seen in clinical trials where patients and/or the doctor don’t know which patients are or aren’t part of the control or test group
What is randomisation?
Where individuals in an experiment are randomly assigned to groups
Why might individuals not be completely randomly put into groups?
Control for block????
Why is randomisation better carried out with a machine than a person?
Introduces bias so may not be truly random
What is the placebo effect?
Where a subject can experience the effects of a treatment/independent variable without it being administered to them