What is the independent variable
The variable that we are in (influencing) the one that we want to see how when levels of it are changed it affects the dependent variable)
What is the dependent variable
the variable who we are trying to see how it is influenced (how it might or might not depend on the independent variable) by different levels of the independent variable
when is the term level vs condition used
to describe different ammounts of the same class of thing - ie different levels of water are given
condition used when different classes of thing are being given (it differs notin the ammount of the same thing being given but rather that it is two different things being given)
ie placebo vs drug condition
What are the extraneous variables
variables outside of the one that we are manipulating
can include random noise
random noise occurs when you do have a variable that is not the manipulation of the independent variable that may impact a specific participants results however we assume that this variable experession is assorted randomly between the two groups (not just in one group and not in the other)
if the groups are divided according to an extraneous variable (systematically - think systemic racism is like assigning people of different races to different living conditions but it is not distributed equally poc more likely to receive specific types of treatment under this- similarly systematic refers to when a specific expression of a variable is not randomly found between groups - is majority present in one group and hardly present in another so will skew the average) we call it a confounding variable bc it confuses our results- we don’t know if the changes we see are due to the manipulation of the independent variable or the confounding variable
How can we try to control for our extraneous variable
Try to have as large a sample size as possible so it should average out (there should be enough people to have each expression of the extraneous varaible in each group)
Try to make as much of what we can the same between the two groups (ie where they take the test, who the test administrator is, etc.)
Try to have nearly identical participants ie 20 year old gay women living in edmonton (con low external validity- results may not apply to most of the poplulation bc some of the population has none of their variables represented here -ie 50 year old gay men not living in edmont)
could also discover a confouding varaible and treat it as an independent variable (so either we are looking at the combo effect of those two variables
Single factor design
single as in we have one indepent variable that is held at 2 different levels or 2 different classes (ie medication no medication)
single factor multi level design
level and condition can both refer to the expxression of the independent variable, level usually means the specific quantity of a variable and conditon usually refers to if the variable is present or a control is used (ie placebo) in a single factor multi level design we have more then 2 different expressions of the independent variable
single factor (non multi level) - just 2 expressions of the independent variable ie one witness vs 2 witness
single factor multi level design - more then 2 expressions of the independent variable
ie one witness, two witnesses or 5 witnesses
treatment condition
here treatment IS defined as something that makes it better
placebo
placebo
placebo effect refers to when people report having their symptoms improve/have thier symptoms appear to improve bc they thought they were recieving treatment. Placebo effect is thought to occur bc when people think they are recieving treatment their stress might be reduced. We want to make sure that if we give a treatment participants improve bc the treatment worked and not bc they were just expecting to improve- solutions around the placebo effect
solutions
make a placebo condition and compare it to treatment condition to gage how much of the impact occurs due to the placebo effect
appears to be identical to treatment condition but does not include the actual treatment (the thing that we think will actually make people improve). ie if it is a drug that is being given the placebo will come in the same pill case but contain sugar or something tasteless that does nothing vs the treatment condition will actually contain the drug. Ie if it is psychotherapy and we think asking specific questions/following a specific structure will make people improve (making this the treatment) then the placebo would be to still given the placebo group psychotherapy but to just talk and avoid adding in what we think will make them improve from the treatment.
wait list condition
control group does not recieve treatment but are told they will be given it right after the trial ends. Since the control group expects to recieve treatment sometime soon this should ease their worries some and therefore this might emulate the placebo effects impact of potentially improving problems by reducing stress.
can also just not have a control condition compare treatment with best avaliable treatment - since we know the best avaliable one works we would see that if the new one is better that it genuinely works bc it shows that it has power beyond placebo effect.
what is a between subject design vs a within subject design
between subject design - each participant is exposed to only one condition
within subject design - each participant experiences all conditions.
between like when you switch between different participants you get different conditions
within you get all conditions within all participants (participants are all exposed to all conditions)
What is the difference between random assignment and random sampling
random assignment - when random processes are used to assign individuals to a condition (needs to occur in an experiment)
random sampling - use random measures to get a group of people from the population (very often not used for experiments - ie we might just be doing surveys - not actually controlling variable levels)
What can be an issue with random assignment and what is a way to solve that
if your doing something like flipping a coin to determine which group participants are in then you could end up having it that you coincidentally flip for heads which is group A a bit more then tails for group B = making it so that group A has a slightly larger sample size. Usually having unequal group sizes is not considered a problem
block random assignment
something generates each condition once - participant list is randomized (and probably the order in which conditons are generated within the parameters that they all have to be generated once). block as in seperate structures (each entire list of conditions is its own strucutre - all conditoins must be listed once in a block (can not be listed twice - bc a block counts as each condition listed once and then move onto another block once you get a repitition) - the order in which conditions are listed within these parameters can vary).
Can also have in theory an issue that random assignment ends up coincidentally dividing participants according to a confounding variable - ie one group has the majority of participants older then the other group
solution - if experiment is replicated nearly impossible this exact issue will happen agin - especially if replication occurs several times
What is a matched group design
We measure participants level of the dependent variable (the variable that we are trying to determine how it will be impacted by modifications to the independent variable) before the experiment and ensure that equal numbers of people with the same level of the dependent variable are put into the two groups
ie if we are trying to see if journalling about traumatic experiences helps people heal which is defined as an improvement to their blood pressure, cortisol levels, etc. We could measure all the peoples blood pressure and cortisol levels in our experiment and then get 1 of the 2 people with the best results to go to the journalling condition and the other to go to the none journalling condition (we don’t choose ourselves ie flip a coin for the first person and then the second person goes to the other condition so we still have random assignment), then do the same for the two people with the second best and go on until you get the people with the worst.
What are the cons of within subjects experiments
between like spread between have conditions spread between the participants - means that you have participant 1 be condition a, participant 2 be condition b and so on
within have all of conditions within each participant ie participant 1 be condition a and condition b, participant 2 is also condition a and condition b
When the order in which the participant was exposed to conditions impacts their responses this is called an order effect
What is better for reducing the impact of noise a within subjects design or a between subjects design
within subjects design bc have each participant experience each of the conditions
What are the different types of order effects
carry over effects
what are carryover effects
carry if have no conditions before the current one have nothing to carry over to it - so no carry over effect can occur unless we have already experienced past conditions
ie practice effect - we carry over our experience from the previous trails making us better at our next trials - so we do the best at the latest point
fatigue effect - we carry over a depletion of energy from our earlier trials so we do the worst on our later trails
context effect /contrast effect - we judge something differently based on how it contrasts with what we have just seen ie ranking attractivness if we have just seen someone attractive we might rank an average looking person more harshly then if we have just seen someone unattractive
What can we do to combat order effects
counter balance
refers to when we ensure that not all participants do things in the same order
difference between a complete and partial counterbalance
when we have equal number of participants doing each possible order.
ie have ABC as our conditions complete counterbalance
ABC
ACB
BCA
BAC
CAB
CBA
not only is there options where A is individually in each possible position (A is in the first space, A is in the second space, and A is in the third space) there is also options for each order that can follow ie if A is in the first space can have B in the second space and C in the third space or B in the third space and C in the second space- in a complete counterbalance we would consider all of these whereas in a partial counterbalance we would not
to calaculate complete counterbalance do n! (n factorial)
n x (n-1) x (n-2), (keep going until you have n number of numbers (after you have calculated the brackets)
ie 3 x 2 x 1
partial counterbalance we are only interested in creating a system where A is in the first space, A is in the second space and A is in the third space. We only choose ABC for the order in which A is in the first space and not ACB - this can be more practical if we have a very large sample size. The number of orders we will use in a partial counterbalance is equal to the number of conditions we have (bc in order for each condition to be in each spot once we need our number of orders to equal our number of spots which = our number of conditions)
random counterbalancing
we determine all the possible different orders by doing n factorial (n!) however it differs from complete counterbalancing in that we don’t have every single one of these orders represented by a participant- instead we randomly choose a set number of orders through random selection (so we will not get a window into all orders just the set number that we have randomly selected) - this is considered less effective then complete or partial counterbalancing
simultaneous within subject design
we mix the order of the conditions within a list (works only if we have multiple examples for each condition) so instead of having all examples for each condition listed at one time and then all the examples for the next condition listed at another time (ie we have all examples for condition A listed and then all examples for condition B) we have them mixed together (ie example for condition A - example for condition B- example for condition A…) - simultaneous bc although we are not seeing each conditions example at the exact same time but we are seeing them more closlely to at the same time more often (in non simultenous design would only see the examples of two different conditions beside each other once - ie when list A ends and list B begins whereas in simultaneous design we would see the examples by each other more often- so it makes it so it is closer to having different conditions shown at once)
what does counterbalancing mean
testing participants in different orders
When should you use a between subjects design vs a within subjects design
Pros of a within subjects design
less impact of extraneous variables
need less participants ( a. bc you don’t need each condition to be represented by a participant and if you do a random counterbalance then you wont do every order of condition so the number of participants can be less then our number of conditions - furthermore for a between subject design it would not be enough to only have one person represent each condition however it would
cons participants might get a better idea of what you are testing bc they know each condition and then can modify their behaviour accordingly
can be more complex to set up and more time consuming so it is not always practical
can have carry over effects
correlation vs causation
correlation - when there is a relationship between two variables- ie when one increases the other either increases or decreases (and this happens enough that we don’t think its chance- represents a pattern). This does not mean that the one variables increase or decrease is causing the other variable to increase or decrease it as they could both have their changes occur due to another variable ie variable 3 if present causes variable 1 to increase and variable 2 to increase - so it is not variable 1 causing variable 2 to increase - they just increase at the same time bc they both increase in the presence of variable 3.
What is the purpose of experiments
first to establish a correlation between two variables (that the variables follow a pattern in what they do in realtion to each other - there is no point in trying to say one thing causes another if they do not follow a pattern and therefore do not appear to be connected) and then to establish a causal relationship (we assume based on the correlation that it is one of the variables we are examining controlling the other that makes this relationship occur and not some other variable and we try to prove this through our experiment)
Internal validity
how valid our conclusions are inside the experiment - ie did we conduct the experiment correctly, did we measure things correctly if not then we can not trust our conclusions if so then gives us more confidence that we can
external validity how valid our conclusion is outside of our experiment (how true this conclusion holds for real life)
What is mundane vs psychological realism
mundane realism - when we are testing a hypothesis and we make the experiment so it resembles the situation in which the individual would encounter the independent variable as defined in the hypothesis in real life
psychological realism - when the experiment mimics the basic psychological processes that might occur when encountering the independent variable in daily life but not the full experience of encountering it in daily life.
ie seeing if people like cereal that is in a blue or purple package more - high mundane realism would be to have people in a fake store and to see if they choose the blue or purple cereal more (this makes it so it is realistic to how they interact with cereal boxes in the real world) - this has high external validity bc closley mimics the situation we are examining
high psychological realism ie have people judge if they like the color blue or purple better to conclude if they will be more likely to buy the purple or blue cereal. This has high psychological realism bc it careful resembles ONE of the psychological processes that is behind their cereal box choice (color preference) however it does not capture all the potential variables that could be present so our conclusions have a low generalizability (a low external validity beyond our situation) - we do not nealry caputre all the factors at play so we might miss some things
