2024-03-11
Great work! I know there was a lot of plotting, but you made some great plots!
I also appreciate your thoughtful approaches to the relationships between variables
Just a reminder: when visualizing, making tables, displaying information from the data: always keep in the back of your mind:
What can a reader get from this if they have never seen the data?
Is it easy for the reader to understand the plot?
Does everything in the plot have a purpose?
Is the main thing I’m trying to communicate also the thing that stands out?
I said glimpse, but that was a bad choice of words
head()
might be better to show each row with their observations and variablesWhen we were looking into observations that might be suspicious…
We should look at the intersection of multiple suspicious observations
A few combinations that people came up with that I thought were good ideas:
Someone who chose that they felt extremely similar to fat and thin people
Someone who reported they were 11-14 years old with education above high school (or other age/education combos)
If the variable is an explanatory variable and originally categorical, it’s good to keep it categorical
lm()
functions with variable treated as continuousSplitting gender identities is NOT mutually exclusive
If you are trying pinpoint one group and make it mutually exclusive, we need to take extra steps
Someone wanted to identify three groups:
Point on data equity:
Always write out all identities within a grouped category
I did not call the group something like non-normative genders, instead I said “Identifies as trans man, trans woman, genderqueer/non-conforming, and/or other”
Your coding names in R
can be different, but when you write it out, make sure you define the group
Example : if someone identifies as a man and a trans man, their data are like:
When hypothesizing whether a variable a confounder or effect modifier
Make sure to back any claims in your final report with sources
We can speculate what’s at play, but we can’t actually know
Our own identity may bias how we perceive specific dynamics
Education: I saw many of us speculate that higher education may be associated with lower IAT (as a potential confounder)
Data showed the opposite: higher education, higher mean IAT score
We can think about people’s perception of controllability of weight: do people assume certain behaviors about fat people?
Does that align with or go against people’s assumptions about behavior needed for higher education?
We also need to think about how education might be linked to socio-economic status, and how that might change what food is affordable
Heavily discussed in Maintenance Phase podcast, but I don’t have direct sources
scale_x_discrete(labels = function(x) str_wrap(x, width = 10))
: use this to wrap the text on x-axis
Keep your explanatory variable on the x-axis when you are plottinf three variables at once
hjust
and vjust
will move your text on the x axis so it does not cover your plot
Plotting age vs. IAT
geom_smooth()
to show moving mean value
Boxplots and plotting each mean not exactly right for continuous variables
See Lab 4 for how I plot this!
In geom_smooth()
, when to use method = lm
If IAT score ranges from -2 to 2, what changes in mean IAT is a lot?
A lot of the coefficients may be significant, but are they clinically meaningful?
Quiz and Lab 2