Biology: Test 3

Front 1

Acceptance of a Model depends on the

1

Back 1

Acceptance of a Model depends on the cost and benefits of accepting it.

Front 2

If a model is REFUTED, it means that it is

2

Back 2

If a model is REFUTED, it means it is rejected. The data so disagrees with a model that we are forced to reject it. Refuting a model is an absolute classification--there is no returning to reconsider a refuted model.

Model: Probability of getting red on a roulette wheel 50 percent of the time.

If you get red 20 times in a row, the model is rejected becuase you have a 100 percent probability, not a 50 percen probability.

Front 3

If a model is CONSISTENT, the data does not

3

Back 3

If a model is CONSISTENT, the data does not cause us to reject it.

Front 4

Two possible reasons for a model being consistent include the scientific definitions of

4

Back 4

Two possible reasons for a model being consistent include the scientific definitions of "support" and "irrelevant."

Front 5

Support means that the data could have

5

Back 5

Support means that the data could have rejected the model but did not, thus causing us to gain some confidence in the model. Supporting a model can be a reversible designation--additional data may ultimately refute the model.

Front 6

Irrelevant means that the data could not possibly have

6

Back 6

Irrelevant meanst that the data could not possibly have rejected the model. This means that it would not matter how the data turned out.

EX: Payoff at a slot machine. Irrelevant, could never reject.

Front 7

Consistent means that the data does not cause us to

7

Back 7

Consistent means that the data does not cause us to reject a model.

EX: First 2 spins on a roulette wheel are red--consistent but not very useful for determining if you will get 20 reds in a row.

Front 8

If a model is to be taken seriously, it must be

8

Back 8

If a model is to be taken seriously, it must be falsifiable.

Front 9

Falsifiable means that we can imagine

9

Back 9

Falsifiable means that we can imagine data that can refute it. If you cannot imagine such data it is usesless. This is a scientific probelm faced by creationist models and intelligent design models.

Front 10

The Two main approaches to deciding what models to use as starting points in a study include the

10

Back 10

The Two main approaches to deciding what models to use as starting points in a study include the "NULL" model approach and the "EQUAL ALTERNATIVES" model approach.

Front 11

The Null model is considered an accepted

11

Back 11

The Null model is considered an accepted model by convention until it is rejected. A Null model is NOT set in stone.

EX. In a coin flip, the probability of heads is 1/2 = NULL Model. In a court trial, the NULL Model is innocent until proven guilty. In Intoxication/Drug Tests, Sober/clean is the Null Model. The FDA rules a substance harmful until proven safe (NULL Model).

Front 12

EQUAL ALTERNATIVES are used when there is no reason to initially believe that

12

Back 12

EQUAL ALTERNATIVES are used when there is no reason to initially believe that one model is more acceptable than the other.

EX: Taste Test of Coca-Cola vs. Pepsi preference model would initially be considered as acceptable as other Coca-Cola preference models.

Front 13

No Data can reject all

13

Back 13

No Data can reject all alternative models. This means that you cannot porve a model.

Front 14

THREE reasons why people continue to disagree with models include the fact that

NESWTSM, ASWFTRLOM, SPCMTCBR

14

Back 14

THREE reasons why people continue to disagree with models include the fact that, 1.) Not everyone starts with the same model, 2.) Any study will fail to reject lots of models, and 3.) Some people choose models that cannot be rejected.

Front 15

The criteria for rejection is never

15

Back 15

The criteria for rejection is never absolute.

Front 16

The Criteria for rejection could occur in the Null Model, as evidenced by the possibilty of

16

Back 16

The Criteria for rejection could occur in the Null Model, as evidenced by the possibility of getting 10 heads in a row. This could happen, but it is NOT likely.

Front 17

The VALUE OF REPEATABILITY ultimately decides which models are

17

Back 17

The VALUE OF REPEATABILITY ultimately decides which models are kept. If the results don't consistently support a model, we will abandon the model.

Front 18

A correlation is similar to a statement of

18

Back 18

A Correlation is similar to a statement of association. For instance, a car that is red is more often in an accident than cars of other colors.

Front 19

Correlation, as it refers to red cars and accident rates, could refer to the driver's

19

Back 19

Correlation, as it refers to red cars and accident rates, could refer to the driver's age, gender, health, height, hair length, etc.

Front 20

Causation implies that X and Y are the reason for an

20

Back 20

Causation implies that X and Y are the reason for an association. For instance "Study finds sex LEADS TO higher drug use." Sex causes people to use more drugs.

Front 21

In Advertising, many ads try to establish an

21

Back 21

In Advertising, many ads try to establish an association between the product and good feelings. (correlation) Beer commercials--"it doesn't get any better than this."

Front 22

If advertising is effective, our associations will lead to a

22

22

Back 22

If advertising is effective, our associations will lead to a causal basis--when I see this product, I'll be happy.

Front 23

Correlation is a pattern of

Causation gives us a reason X and Y are

23

Back 23

Correlation is a pattern of association. Causation gives us a reason X and Y are correlated.

Front 24

Causation gives us a reason to believe that if X causes Y and we change X, then

24

Back 24

Causation gives us a reason to believe that if X causes Y and we change X, then we will also get a change in Y.

Front 25

An example of a CAUSAL MODEL would be a model that says smoking causes

25

Back 25

An example of a CAUSAL MODEL would be a model that says smoking causes lung cancer: meaning that if we smoke, our chances of getting lung cancer goes up.

Front 26

An example of a CORRELATIONAL MODEL would be a model that says smokers have a higher rate of

26

Back 26

An example of a CORRELATIONAL MODEL would be a model that says smokers hava a higher rate of lung cancer than non-smokers.

Front 27

Another CAUSAL MODEL would be a model that says talking on a cell phone while driving causes

27

Back 27

Another CAUSAL MODEL would be a model that says talking on a cell phone causes drivers to have higher accident rates.

Front 28

Another CORRELATIONAL MODEL would be that drivers talking on cell phones have higher accident

28

Back 28

Another CORRELATIONAL MODEL would be that drivers talking on cell phones have higher accident rates than drivers NOT talking on cell phones.

Front 29

We usually want a causal model, because understanding causation tells us what to

29

Back 29

We usually want a causal model, because understanding causation tells us what to change to achieve our goals.

Front 30

Correlation does NOT IMPLY

30

Back 30

Correlation does NOT IMPLY CAUSATION.

Front 31

You cannot infer CAUSATION from CORRELATION because of

31

Back 31

You cannot infer CAUSATION from CORRELATION because of HIDDEN VARIABLES.

Front 32

Hidden Variables could explain why red cars have the

32

Back 32

Hidden Variables could explain why red cars have the highest accident rates.

RED CARS--Risky Drivers
RED CARS--Risky Cars
RED CARS--Old Car

(Hidden Variables)
OTHER COLORS--Safe Drivers
OTHER COLORS--Safe Cars
OTHER COLORS--Young Car

Front 33

Hidden Variables are also known as

33

Back 33

Hidden Variables are also known as Third Variables.

EX. Causal Model--Risky Cars tend to be painte red. Cause--Driver. Third Variable--Car Type.

If we only look at X and Y, then we cannot see the effect of the third varible, Z.

Front 34

The solution to the THIRD VARIABLE problem is to

34

Back 34

The solution to the THIRD VARIABLE problem is to control for hidden variables.

Front 35

One way to control hidden variables is to control for the
35

Back 35

One way to control hidden variables is control for the Third Variable by design.

Front 36

Another way to control hidden variables is to do an

36

Back 36

Another way to control hidden variables is to do an experiment that automatically controls for hidden variables.

Front 37

To CONTROL for a variable is to make it the

37

Back 37

To CONTROL for a variable is to make it the same across the different groups we are studying. (on average)

Front 38

Making a variable the same could include making it

AIAGS, PIAGS, MIAGS

38

Back 38

Making a variable the same could include making the variable absent in all groups studied, present in all groups studied, or mixed equally in all groups studied.

Front 39

To control for a car type in the red car study, a specialized

39

Back 39

To control for a car type in the red car study, a specialized box is used showing red sports car, red safe car, non-red sports car, non-red safe car.

Front 40

The goal of using this specialized box is to

40

Back 40

The goal of using this specialized box is to eliminate/control for as many variables as we can.

Front 41

The Box approach to testing to testing apart correlation and causation is limited because we can only

41

Back 41

The Box approach to testing apart correlation and causation is limited because we can only attempt to use factors we can think of (what about car's age, for example.) Also, it may difficult to get a sample that will allow us to control for the factors we think of.

Front 42

The Models of Intelligence used in the Monty Python Bird Video are

42

Back 42

The Models of Intelligence used in the Monty Python Bird Video are Brain Size and IQ Score/Test Performance.

Front 43

Brain Size was confounded with ___ ___ as a third variable.

43

Back 43

Brain Size was confounded with body size as a third variable.

Front 44

To control for body size, the scientists figured out what the brain size would be for a penguin as tall as a

44

Back 44

To control for body size, the scientists figured out what the brain size would be for a penguin as tall as a person. This would equal RELATIVE BRAIN SIZE.

Front 45

Test Performance was initially confounded by the the

45

Back 45

Test Performance was initially confounded by the testing environment and language.

Front 46

How could we control for everything at once? If we could randomly assign

46

Back 46

How could we control for everything at once? If we could randomly assign car color, leaving everything else the same, we would now have equalized all other 3rd variables across car color.

Front 47

If we were to randomly assign car color and leave everything else the same, this would be an example of a

47

Back 47

If we were to randomly assign car color and leave everything else the same, this would be an example of a planned manipulation to test a model.

Front 48

Correlational Data do not involve

48

Back 48

Correlational Data do not involve manipulations (changing the natural order.)

Front 49

Experiments are a planned

49

Back 49

Experiments are a planned manipulation to test a model.

Front 50

One cause of unwanted correlations deals with us knowing what the

50

Back 50

One cause of unwanted correlations deals with us knowing what the variables are and creating and creating an experiment to separate the two.

Suppose your mechanic changes your spark plugs and tells you to switch to high octane fuel. If you do this, your mileage improves. Is it the plugs, the fuel, or both that are causing you to get better gas mileage? Using the box method, there are 4 different ways control for spark plug type and fuel type.

Front 51

Another case of an unwanted correlation involves us now knowing what the

51

Back 51

Another case of an unwanted correlation involves us not knowing what the variables are.

Front 52

To resolve the problem of not knowing what the variables are, we must randomize the assignment of

52

Back 52

To resolve the problem of not knowing what the variables are, we must randomize the assignment of treatment versus the control group to destroy all unwanted correlation.

Front 53

If a model makes use of two models of the same car, both differing in the use of color (red, non-red), and uses the same driver, what is the control in the experiment?

53

Back 53

Car color. However, time of driving and other possibilities are not controlled for.

Front 54

If you take cars as they exist and randomize the drivers across the cars, what is controlled for?

54

Back 54

The driver characteristics are controlled for.

Front 55

What if you take the car-driver combination as they exist and randomize color?

55

Back 55

If you take the car-driver combination and randomize color, you have controlled for all hidden variables.

Front 56

All issues of ideal data such as ___, ___, ___, and ___ apply to

56

Back 56

All issues of ideal data such as blind, standards, replication, and controls apply to experiments.

Front 57

When trying to determine the truthfulness of autistic children using facilitative-communication, the two things researchers looked for regarding allegations were:

57

Back 57

When trying to determine the truthfulness of autistic children using facilitative communication, the two things researchers looked for regarding allegations were, (1) Are the kids lying about sexual abuse? and (2) Are the words the kids write really their own words--not the words of the facilitator?

Front 58

To determine the effectiveness of facilitative communication, researchers gave both the facilitator and the autistic child

58

Back 58

To determine the effectiveness of facilitative communication, researchers gave both the facilitator and the autistic child blinded tests. The facilitator would be given a picture that the autistic child could see and the autistic child would be given a different picture that only he or she could see. If the child typed what the facilitator saw, then researchers knew that the autistic children were simply being manipulated by the facilitator.

Front 59

A Null Model is a default

59

Back 59

A Null Model is a default model--one chosen just to have an obvious starting point.

Front 60

A Null Model could be a model that we think is commonly

60

Back 60

A Null Model could be a model that we think is commonly true or a model that we don't think is true. A Null Model is used to demonstrate that we can reject something. (Coin Flips, alcohol impairment, etc.)

Front 61

The inability to reject all possible alternatives is the

61

Back 61

The inablity to reject all possible alternatives is the main reason we can never prove that a model is correct.

Front 62

The most common criteria for the acceptanc of/rejection of a model is based on

62

Back 62

The most common criteria for the acceptance of/rejection of a model is based on statistics.

Front 63

Statistical tests are mathematical tools that tell us how often a set of data is

63

Back 63

Statistical tests are mathematical tools that tell us how often a set of data is expected by chance under a particular model.

Front 64

If a set of observations could be expected to occur under a particular model only

64

Back 64

If a set of observations could be expected to occur under a particular model only 1 in 20 times (5%), we reject the model.

Front 65

Repeated Successes cause models to be

65

Back 65

Repeated Successes cause models to be accepted because they have been proven to work time and time again.

Front 66

Random is reserved to explain why we get different

66

Back 66

Random is reserved to explain why we get different outcomes when we keep everything the same. (Coin Flip--two possible outcomes with equal probability.)

Front 67

In Randomness, a statistical test replicates the model of

67

Back 67

In Randomness, a statistical test replicates the model of randomness to see how often the random process fits the real data.

Front 68

The chance of two unrelated people having the same birthday is approximately

68

Back 68

The chance of two unrelated people haeving the same birthday is approximately 1 in 365. In a group of 23, the chances that two people share the same birthday is approximately 50%.

Front 69

The reason for the paradox between 1 in 365 and 2 in 23 is due to

69

Back 69

The reason for the paradox between 1 in 365 and 2 in 23 is due to the fact that there are many different pairs of individuals to consider in a group of 23 (253 pairs to be exact), although not all pairs are ‘independent’ of the others.

Front 70

Variables are things we measure that can differ

70

Back 70

Variables are things we measure that can differ from one observation to the next, such as height, weight, behavior, fat intake, life-span, grade-point average, and income.

Front 71

In general, a variable is a measure of something that can

71

Back 71

In general, a variable is a measure of something that can take on more than one value

Front 72

When an association exists between two variables, it means that the average value of

72

Back 72

When an association exists between two variables, it means that the average value of one variable changes as we change the value of the other variable

Front 73

When a correlation is weak, it means that the average value of one variable changes

73

Back 73

When a correlation is weak (e.g., Model C), it means that the average value of one variable changes only slightly (only occasionally) in response to changes in the other variable

Front 74

When a correlation is a strong-positive, the left end of the line on a graph is significantly

74

Back 74

When a correlation is strong, the left end of the line on a graph is signifcantly lower than the right end of the line.

Front 75

When a correlation is a strong-negative, the left end of the line on a graph is significantly

75

Back 75

When a correlation is a strong-negative, the left end of the line on a graph is significantly higher than the right end of the line.

Front 76

If the points on a graph pretty much fall inside a circle or horizontal ellipse so that the "trend-line" through them is horizontal, then a correlation

76

Back 76

If the points on a graph pretty much fall inside a circle or horizontal ellipse sso that the "trend-line" through them is horizontal, then a correlation does not exist, the same as a zero or no correlation graph.

Front 77

When either or both variables cannot be assigned numbers a correlation may still exist but we no longer

77

Back 77

When either or both variables cannot be assigned numbers, a correlation may still exist but we no longer apply the terms positive and negative. Strong is the only term used.

Front 78

A control serves as a reference

78

Back 78

A control serves as a reference point for the study, i.e., a point of comparison.

Front 79

The control is possibly the most vital design feature in studies testing

79

Back 79

The control is possibly the most vital design feature in studies testing causal models (or other models which make a comparison).

Front 80

A causal model can be evaluated with a set of data only if the data, MTRCDBTM, and TDITBTGMTCGITM

Back 80

A causal model can be evaluated with a set of data only if the data measure the relevant characteristics described by the model, and
(ii) the difference in treatment between the groups matches the comparison given in the model.

Front 81

The problem with correlational data is that one often does not know how many factors differ between the

81

Back 81

The problem with correlational data is that one often does not know how many factors differ between the main group (treatment group) and the control group.

Front 82

factor can be controlled for if

i) IIAITACG, ii) IATEITCATG, iii, IPIOSMOEGBIPTTSDBCATG

82

Back 82

factor can be controlled for if

(i) it is absent in the treatment and control groups,

(ii) it applies to everyone in the control and treatment groups, or

(iii) is present in only some members of each group but is present to the same degree between control and treatment groups.