PSYC 412 Lecture Notes - Lecture 6: Statistical Inference, Developmental Psychopathology, Null Hypothesis
14 May 2018
School
Department
Course
Professor
PSYC 412: Developmental Psychopathology
Jan 10th, 2018
Lecture 2: Stats for Success
STATISTICS
• Descriptive vs. Inferential statistics
o Descriptive stats are used to describe a data set
▪ What is the middle, central tendency of a set of numbers?
▪ Central Tendency
• Mean - average
• Median – indicator of central tendency, score where there are
equal numbers of people before and after
• Mode – most frequently appeared score
▪ Measures of spread
• Variance
o Is there a big spread in the data?
• Standard deviation
o Also measures data spread
▪ Useful for describing data sets: the first thing done during research
projects when analyzing initial data results
▪ Limited
• Descriptive stats describe the data set that you
have
but
doesn’t allow you to talk about people not included in the data
set
• i.e.: you wanted to know average GPA of U3 psych majors
o Not possible to ask every single U3 psych major so you
pull data from 20 people as proxy for everybody
o Calculate mean for those 20 people, but it is not clear
that this mean reflects entire population of U3 psych
students, and this is why we need inferential statistics:
they let us think about world beyond data we collected
• Sample: group actually studied
• Population: group you are really interested in
find more resources at oneclass.com
find more resources at oneclass.com
o Inferential statistics
▪ Sample vs. population
• Population is all the U3 psych students (from above example)
• Sample is the small group of 20 people
▪ Ex: do U3 psych majors at McGill have higher GPAs than U3 economic
majors?
• Obtain GPA of every single U3 econ student and every single
U3 psych major – then you can literally just compare the means
▪ What if the small group of 20 psych students happened to be very
high than average, and the econ group was much lower than average?
Could not be very reflective of what’s going on in the population: you
need a way to index this
• Inferential stats gives you this capability.
▪ Example:
• Psych students mean GPA is 3.0, Econ mean GPA 2.5
o Is the difference we observe between the two groups
dependable or did we just get this information by
chance?
o Hypothesis testing allows you to answer this question
• Null hypothesis: there is no difference between these two
groups
o Econ majors and psych majors have the same GPAs^
• P-value: probability of obtaining a difference as big as the one
I got if there actually is no real difference. They have the same
GPAs, what is probability that I obtained the difference of 0.5
points between GPA of the two groups
by chance
?
o Typically .05
o When p < .05 there is less than 5% chance that we
would have obtained a difference that big if in reality no
difference existed.
o If p is greater than .05 then we aren’t sure if this
difference really reflects what’s happening in the
underlying population
o What types of statistical tests do developmental psychopathologists use?
find more resources at oneclass.com
find more resources at oneclass.com
▪ Often interested in differences between groups.
• Does a treatment group (kids who got treatment for a
psychological disorder) have fewer symptoms than a control
group?
• Conduct problems=poorer language skills than those without
conduct groups?
▪ Are continuous variables related to each other?
• No longer breaking into groups, but looking at continuous
scores.
• Ex: is lower family income associated with increased conduct
problems?
o No groups, just changes in family income
• Ex: is greater age associated with increased depressive
symptoms?
o No groups, continuous spectrum of increasing age
• Relationship between two continuous variables: statistical
symbol is r and is called a correlation
o Positive correlation: as the independent variable
increases, the dependent variable also increases
▪ I.e. as age goes up, depressive symptoms go up
with it
▪ Bounded by -1 and 1, largest value for a pos corr
is 1 = perfect association between two variables
▪ Correlation of 0.3 is a very healthy relationship in
this world (considered a good size association)
o Negative correlation: as the independent variable
increases, the dependent variable decreases
▪ i.e. as family income goes up, conduct disorders
go down
▪ -1 is perfect negative correlation
o No correlation: value is 0, no relationship between the
two variables, looks like a random scatter plot
o What if I have more than two variables?
▪ You always will have >2 variables
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Could not be very reflective of what"s going on in the population: you need a way to index this. Inferential stats gives you this capability: example, psych students mean gpa is 3. 0, econ mean gpa 2. 5. I got if there actually is no real difference. Dvs): don"t skip this part, then think about what the main effects and the interactions are. Methods: research methods i, basic review of research methods, nosology: categorization in developmental psychopathology, categorical vs. dimensional measurement. This is present for everyone in varying degrees: some are higher than others but are not fundamentally different from other people. Issues in measurement: how do i measure something, what am i interested in studying, you need to define your construct. I am actually measuring what i think i"m measuring. Inter-rater reliability: there is agreement between two people in judging whether something is present or occurring, diagnosis, example: can two clinicians agree that a child has.
