research methods

Question 1

What are quantitative methods?

Answer 1

Measurements (numeric, quantification)

Using scales (e.g. for pain but can be problematic in humans as subjective)

induces PREDICTABILITY - this is often the goal of science (allows survival)

makes science apolitical, valueless and unbiased

Question 2

What are qualitative methods?

Answer 2

Describing, understanding a phenomena.

Understanding what it’s like to be or to have.

The EXPERIENCE

This is also SUBJECTIVE but qualitative takes much more account of individual perceptions than quantitative

Question 3

What are the levels of measurement?

Answer 3

(LOW/MINIMUM/DISCRETE)
1. Nominal
2. Ordinal
3. Interval
4. Ratio
(HIGH/MAXIMUM/CONTINUOUS)

Only higher levels allow prediction

Question 4

What is NOMINAL measurement?

Answer 4

1 (lowest)

categorical, but sometimes no categorical/quantitative difference between categories. Only allows for descriptive stats

Question 5

What is ORDINAL measurement?

Answer 5

2

still categorical but starting to be able to say one is bigger than another eg political voting behaviour. There may not be units between points.

Question 6

What is INTERVAL measurement?

Answer 6

3

no zero eg age or height which we can measure and find infinite divisions and we know there is always the same difference between intervals (in a cm for example)

Question 7

What is RATIO measurement?

Answer 7

4 (highest)

a continuous scale where there is an absolute zero e.g. velocity.

Question 8

About qualitative types of procedure

Answer 8

Historically, qualitative information has been missed.

Observation - ask them to describe their experience eg with colorectal cancer (very common, 3rd most
Likely cause of death with ~10,000 deaths in UK per year).

–> If we can understand what it’s like to have something then their care can be improved - what they would have liked to happen.

Want as big n as possible in quantitative BUT in qualitative we want a representative but not too huge as analysis takes a very long time –> 6-20 subjects average for qualitative

Science protects objectivity. Science sometimes finds answers which aren’t pleasant or beautiful (eg evolution) - people don’t always like it.

Science often wants truth which can be unpleasant - eg a diagnosis (colorectal cancer!) - but this is a truth.

Humanity has sought to protect this method/science –> apolitical, valueless, unbiased.

For qualitative research:

- Need 6-20 subjects
- Can conduct interviews
- Input is always verbal or textual (interview is verbal [recordings, transcription, thematic analysis], diaries or facebook is textual)

Question 9

What are types of qualitative methods?

Answer 9

• discourse analysis
• thematic content analysis
• grounded theory

Question 10

What is discourse analysis?

Answer 10

Qualitative method

Much more attentive to the type of language they used than the meaning of it. How they use language is very important and it is thought to say something about their experiences and beliefs - eg swearwords in languages and cultures (fewest in Japanese and most in Russian).
WHY does this vary culturally? Believed that language reflects the culture - needed to develop a language to reflect their experience - “language encodes culture”

Question 11

What is thematic content analysis?

Answer 11

Qualitative method

This is the process of extracting themes and immersing oneself in the data trying to understand as if it were yourself whilst trying not to bring bias and also you are not them so this is very hard. They are a different person with different subjective views, which can never be overcome. No matter what you do no one will ever fully understand you. Psychologists have said because of this that you are truly alone from other subjects and objects. One day we may be able to upload our self and our experiences to some kind of machine so it wouldn’t be lost when we die and could be explored by others. We can do this yet so instead we do qualitative research.

If 200-300 themes are extracted from data then we look for the common recurring ones until we have 5-10 overarching themes which encompass all of the 300 themes. Then a short paragraph would be constructed encapsulating what each theme is/means, with as much of what they actually said as possible. We can quantify the ones which are the most mentioned - what’s not there is as important as what is there.

Question 12

What is grounded theory?

Answer 12

Qualitative method

Start with theoretical premise which is dangerous in qualitative because you may interpret their subjectivity through your theoretical framework eg holding doors - kind or sexist? A priori theory is where you make a theory before hand and this takes certain stand points eg what it’s like to be a woman in the 21st century.

Question 13

What is scientific tenure?

Answer 13

Tenure was originally developed to allow scientists to ask these questions without consequences in order to protect science.

Question 14

What are the types of statistics?

Answer 14

1. descriptive stats
2. non-parametric stats
3. parametric stats

Question 15

What are examples of descriptive stats?

Answer 15

graphs, bar charts, mean, mode, median

Question 16

Why are non-parametric stats?

Answer 16

can be used if data can be ranked but is less accurate (e.g. chi squared)

Question 17

What are parametric stats?

Answer 17

T-tests, ANOVE (analysis of variance - amount of variability within your data)
Eg mean IQ in males vs females. Mean IQ globally for both makes and females is 1–, but really doesn’t stand world wide as female education is rare in many places. If it is measured in the right way in the right places then you can get the same value.

Question 18

What does a T-test show?

Answer 18

• Nothing to do with relationship
  
  • To do with variance within 2 groups can not do it with more than 2 groups
  • Tells you if there’s a significant different between the two values
  • Continuous data
  • Looking for a difference in the score on the two groups
  • Eg statistically comparing males and females IQ

P value = 0.05 –> means that if the value is below 0.05 there is a 95% certainty that the results are not due to chance. If the value is set at 0.01 then there is a 99% certainty.

Question 19

What are experiments?

Answer 19

Experiment = GOLD STANDARD!

1. Manipulation of one variable = independent variable
2. Look for effect on dependent variable - continuous
3. Random allocation to groups

Finds CAUSALITY which is the only way can establish predictability. Psychiatrically we cant give someone a disorder so can’t really do experiments, so cant really establish causation. To get around this we use correlational data. This can’t establish causality. Instead finds a linear relationship between two variables - cant establish causality with this.

When there is a linear relationship between 2 variables we like to say that one causes the other.

Question 20

What are the problems with correlation?

Answer 20

1. We like to infer causation because this allows predictability (previously would help survival)
   
   2. 3rd variable problem - when another variable that you’re not measuring causes both fo the other things - eg hot weather on ice cream sales and drowning
      
      • Partial correlation = correlation between 2 variables while controlling for another variable. If you take out the third variable then the one youre controlling for then the correlation drops to ZERO and this shows the relationship between the two isnt real and was caused entirely by the third variable - a spurious relationship.

Measurements at one time point = cross sectional
Measurements at several time points = longitudinal or cohort

Question 21

How might experiments be done in a drug trial?

Answer 21

Three groups (control, old drug, new drug) given IQ test and measure how long they take to solve it to find speed of processing which is the dependent variable. One month after the drug is given you would retest and compare the mean of the two set of results. We would predict that the control group stays the same, the rug currently used would have some effect and the new proposed drug would have more of an effect than that.

To test for significance we use analysis of variants = for more than two groups
If three groups, for example, then the results of analysis of variance gives an f value which states whether there is significance at p=0.05. The problem is that it doesn’t show WHICH groups are significant from which others.

We then do post-hoc tests.

Question 22

What are post-hoc tests?

Answer 22

These are only done if significance is found - otherwise it isn’t worth it. These are types of t-tests between the groups, between each possible pairing of the three groups.

In total at this point you would have done 3 tests –> CAPITALISATION ON CHANCE which says that the more tests you do, the more likely you are to find something significant by chance. This has to be corrected for statistically by setting the p value lower making it harder for something to be accepted as significant.

Question 23

What is the Bonferroni correction?

Answer 23

• Work out how many tests you’re doing (eg = 4), and you want the p value to be 0.05
  
  • Divide the p vale by the number of tests
  • Eg 0.05/4 = 0.0125 0.013 - which is now your new p value

This is how you account for capitalisation on chance. The problem with this however is that it will approach zero (0.001) with the more tests you do, eventually making it near impossible to find anything significant.

Question 24

What is multiple regression?

Answer 24

If we want to predict a variable fro 2 other variables we do this by multiple regression. We predict one variable from several others - needs interval or ratio level of measurement. Gives three axes with a 3D plane on the graph. (A line of best fit on a scatter grpah shows correlation aka simple regression).

N dimensional space - can be 5 places.

Multiple regression allows us to make predictions on things such as likelihood of getting certain diseases.

Question 25

What is parsimony?

Answer 25

Aparsimoniousmodel is a model that accomplishes a desired level of explanation or prediction with as few predictor variables as possible. For model evaluation there are different methods depending on what you want to know.
–> The simplest explanation is the best.