Behavioral research methods Lectures
Lecture 1
Kinds of variables:
- Categorical / nominal
o Two or more categories without intrinsic ordering (kind of movie)
o When only two categories, also called binary variable or dummy variable
(gender)
- Ordinal
o Two or more categories with intrinsic ordering (5 point rating)
- Interval / continuous
o Ordinal + intervals between values are evenly spaces (age, income)
Y=
- Dependent variable
- Response variable
- Y- variable
- Explanandum
- Target variable
X=
- Independent variable
- X – variable
- Predictor variable
- Explanans
Experiment : researcher determines X Manipulate
Survey : researcher measures X measure
Dealing with data: general analysis setup
1. Check your data
a. To get acquainted with it
b. For outliers and coding errors
2. Determine the kind of analysis (X Y)
3. Recode your data so that you have all variables in appropriate format, create new
variables from existing ones
4. Check the assumptions for the analysis of choice
5. Run your analysis
6. Check the rest of the assumptions
Lecture 2
Statistical inference
= concluding something about the population based on a sample
Procedure of hypothesis testing
1. Define H0 (this involves an equality, and from this, the Ha, immediately follows, that
is its complement)
2. Check the data, and calculate or simulate how likely it is to end up with your data if
H0 would be true (typically stat does this for you)
3. Draw a conclusion
4. Confidence intervals are sets of values for which H0 is not rejected
,Calculating how likely data is “under H0”
We need to know the distribution of this statistic, under H0, to know how likely it is that out
data is drawn from this distribution
Empirical distribution: the center
- Measures for the center of a distribution
o Median : 50% more and 50% less
o Mode : occurs most
o Mean favorite
- Measures for the spread of the distribution
o Max – Min
o P75 – p25
o Standard deviation
Empirical distributions can take on any kind of shape.
We can measure the shape: skewness and kurtosis
The best know theoretical distribution: the standard normal distribution
- Has a mean mu and standard deviation sigma
Hypothesis testing terminology:
- H0 = the baseline of hypothesis
- Alpha = the probability of rejecting H0, when it is actually true (“how likely is it, if H0 is
true, that I get data as I have them, or data that are further away from H0?”)
- P-value < Alpha H0 rejected
- H0 is always supported never true and rejected and not false
The central limit theorem
- The mean of a large enough sample from an arbitrary distribution has a shape of a
normal distribution
if p > 0.05 then H0 is supported, if p < 0.05 then H0 is rejected.
, Lecture 3 calculating scales
Item battery / scale
Item batteries
- Single question is a rough scale
- Item value consists of
o True value
o Item-specific value
o Noise
- With more items, noise cancels out
- With more items, distribution will be more normally distributed
Generating scale score
“alpha feeling*, gen (happy) item
Hist happy “
Scale score = mean of items (after reversing)
- Stata automatically transforms negative items, by adding a minus
- Scale is not from 1-6 any more
“replace feeling1 = 7 – feeling1
Replace feeling4 = 7 – feeling4
Alpha feeling*, gen (happy2) item
Hist happy2”
Recode = 1 specific value and replace = the whole row will be changed
Points of warning
- Scale score – mean of items
- Stata automatically determines a sign
- If you do not want this use
o “alpha varlist, gen (newvar) item asis”
Points of warning 2
- If almost all values are missing, a mean might be based on a single observation
- To ensure at leas 5 observations use:
o “alpha varlist, gen(newvar) item min (5)”
Points of warning 3
- If the values are not on the same scale, this will be wrong
- To ensure that variables have the same scale
o “alpha varlist, gen (newvar)std
Lecture 1
Kinds of variables:
- Categorical / nominal
o Two or more categories without intrinsic ordering (kind of movie)
o When only two categories, also called binary variable or dummy variable
(gender)
- Ordinal
o Two or more categories with intrinsic ordering (5 point rating)
- Interval / continuous
o Ordinal + intervals between values are evenly spaces (age, income)
Y=
- Dependent variable
- Response variable
- Y- variable
- Explanandum
- Target variable
X=
- Independent variable
- X – variable
- Predictor variable
- Explanans
Experiment : researcher determines X Manipulate
Survey : researcher measures X measure
Dealing with data: general analysis setup
1. Check your data
a. To get acquainted with it
b. For outliers and coding errors
2. Determine the kind of analysis (X Y)
3. Recode your data so that you have all variables in appropriate format, create new
variables from existing ones
4. Check the assumptions for the analysis of choice
5. Run your analysis
6. Check the rest of the assumptions
Lecture 2
Statistical inference
= concluding something about the population based on a sample
Procedure of hypothesis testing
1. Define H0 (this involves an equality, and from this, the Ha, immediately follows, that
is its complement)
2. Check the data, and calculate or simulate how likely it is to end up with your data if
H0 would be true (typically stat does this for you)
3. Draw a conclusion
4. Confidence intervals are sets of values for which H0 is not rejected
,Calculating how likely data is “under H0”
We need to know the distribution of this statistic, under H0, to know how likely it is that out
data is drawn from this distribution
Empirical distribution: the center
- Measures for the center of a distribution
o Median : 50% more and 50% less
o Mode : occurs most
o Mean favorite
- Measures for the spread of the distribution
o Max – Min
o P75 – p25
o Standard deviation
Empirical distributions can take on any kind of shape.
We can measure the shape: skewness and kurtosis
The best know theoretical distribution: the standard normal distribution
- Has a mean mu and standard deviation sigma
Hypothesis testing terminology:
- H0 = the baseline of hypothesis
- Alpha = the probability of rejecting H0, when it is actually true (“how likely is it, if H0 is
true, that I get data as I have them, or data that are further away from H0?”)
- P-value < Alpha H0 rejected
- H0 is always supported never true and rejected and not false
The central limit theorem
- The mean of a large enough sample from an arbitrary distribution has a shape of a
normal distribution
if p > 0.05 then H0 is supported, if p < 0.05 then H0 is rejected.
, Lecture 3 calculating scales
Item battery / scale
Item batteries
- Single question is a rough scale
- Item value consists of
o True value
o Item-specific value
o Noise
- With more items, noise cancels out
- With more items, distribution will be more normally distributed
Generating scale score
“alpha feeling*, gen (happy) item
Hist happy “
Scale score = mean of items (after reversing)
- Stata automatically transforms negative items, by adding a minus
- Scale is not from 1-6 any more
“replace feeling1 = 7 – feeling1
Replace feeling4 = 7 – feeling4
Alpha feeling*, gen (happy2) item
Hist happy2”
Recode = 1 specific value and replace = the whole row will be changed
Points of warning
- Scale score – mean of items
- Stata automatically determines a sign
- If you do not want this use
o “alpha varlist, gen (newvar) item asis”
Points of warning 2
- If almost all values are missing, a mean might be based on a single observation
- To ensure at leas 5 observations use:
o “alpha varlist, gen(newvar) item min (5)”
Points of warning 3
- If the values are not on the same scale, this will be wrong
- To ensure that variables have the same scale
o “alpha varlist, gen (newvar)std
