Quiz 3 (5a, 5b, 6a) Flashcards
linkage disequilibrium: sections of DNA may __. alleles at nearby SNPs may __. in other words, if you know the genotype at one SNP, you can __
be inherited together; together more often than you would expect based on their allele frequencies; predict the genotype at a nearby SNP pretty well
humans have __ chromosomes (__ pairs of __ and __ __ chromosomes) in each cell
46; 22 pairs of autosomes; 2 sex chromosomes
meiosis is a type of __ that results in __
cell division; reproductive cells
haploid DNA (__) ends up in germ cells (__ and __)
one set per cell; sperm and oocyte
an autosome is __
any chromosome that isn’t a sex chromosome
when __ occurs, the double helix is broken in __ and __ in homologous places and the ends __
cross-over; one maternal and one paternal chromatid; are combined to form new chromatids that are a combo of DNA from each parent
the homologues have __ but are __
the same type of info; from different parents
a chromatid is __
one half of 2 identical copies of a chromosome
DNA is replicated before __ resulting in __ which are 2 identical copies of a single chromosome (from one parent) that are __
cell division; sister chromatids; connected by a centromere
chiasma =
crossing over point
oocyte =
immature ovum (egg cell)
crossovers occur __ per chromosome, and occur __
many times; anywhere
variants close to each other tend to have __ crossovers between them and so their alleles are __
fewer; more correlated
when the alleles at 2 SNPs are correlated, __
those SNPs are in linkage disequilibrium
ex: usually when SNP1 is an A, SNP2 on the same chromosome is a T. This indicates that this portion of the DNA __
tends to be inherited together/ is in high linkage disequilibrium
if a measured trait is associated with a particular variant, the trait may be cause by __
a different variant that is in high LD with the SNP you measured
__ and __ are two measures of LD
r^2 and D’
r^2 and D’: 0 values indicate __
no LD
r^2 and D’: 1 values indicate __
perfect LD
(r^2/D’) __ is sensitive to minor allele frequency but __ is not
r^2; D’
r^2 tells you __
how correlated the alleles are between two SNPs
D’ tells you __
how often your minor allele appears with a particular allele in another SNP
if you have a high __ but a very low __ you may not want one SNP substituting for the other because this means that __ predicts __ but the opposite is not true
D’; r^2; minor allele in SNP2; minor allele in SNP1
no clear-cut rules for what makes a strong or weak LD:
strong LD: __
intermediate: __
weak/no LD: __
r^2 >/= 0.8;
0.1 = r^2 < 0.8
r^2 < 0.1
__: a factor that can have multiple possible values
variable
__: the outcome whose variability is being measured
dependent variable
__: the variable we hypothesize will explain some part of the dependent variable’s variability
independent variable
“we hypothesize that the variability in the __ depends on, or can be explained by, the __”
dependent variable; independent variable
the probability that you are willing to accept that something due to random chance will be accepted as a true association
alpha
the probability of getting a result that is as strong/stronger than yours if there is no true effect or association
p-value
p-value is usually set to __ for scientific studies
0.05
p-value of 0.05: “we accept a probability of __ suggesting a relationship is real when it __”
<5%; was due to random chance
slide 17?
.
the center value of a defined set of numbers
mean/average
why do we care about mean? lets you __ at a glance, although it doesn’t tell you __
compare group differences; whether those differences are significant
u = (x1 + x2 + x3 +...xn)/n u = \_\_ x = \_\_ n = \_\_
mean; individual value in the data set; number of items in the data set
a measure of how spread out the data points are from the mean
variance
s^2 = z ….
.
standard deviation is __
the square root of the variance
continuous variables: variable with __ ex: __
an infinite number of possible values between two points; hippocampal volume
categorical variables (3)
dichotomous, ordinal, nominal
dichotomous: variables w/ __ ex: __
only 2 levels; does the subject live w/in 5 miles of a highway? (any yes/no question)
ordinal: variables w/ __ ex: __ or __
at least 2 categories that can be ranked; APOE genotype; past smoking behavior (never, <10 years, or >10 years)
nominal: variables w/ __ ex: __
at least 2 categories that DON’T have a particular order; data acquisition site
common stats tests (6)
chi-squared test, student’s T-test, ANOVA, correlation/linear regression, multiple regression, logistic regression
With a nominal variable, the data in each category might be __, but there is no __ (unlike something like __ where __)
different; intrinsic order to those categories; APOE genotype; risk increases with each e4 allele
chi-square goodness of fit: use when you want to determine __
whether the number of subjects in each category of a nominal variable fits an expectation (ex: are the number of people with each genotype different from what we would expect given the minor allele frequency?)
chi-square test of independence: use when you have __ and you want to determine __
two nominal variables; whether the proportions for one variable are different for different levels of the other variable (ex: sex & disease, is the disease more common in males than in females?)
what is a significant chi-square test?
…
student’s t-test: use when you want to determine __
whether a continuous variable is different in two groups (ex: is hippocampal volume smaller in APOE4 carriers versus non-carriers?)
what stats test is this an example of: is hippocampal volume smaller in APOE4 carriers versus non-carriers?
student’s t-test
what stats test is this an example of: variables = sex, disease; is the disease more common in males than in females?
chi-square test of independence
what stats test is this an example of: are the number of people with each genotype different from what we would expect given the minor allele frequency?
chi-square goodness of fit
ANOVA: if you want to covary for other measures (use covariates), use __ (__)
ANCOVA (analysis of covariance)
ANOVA (1-way) aka __. use to measure __. will tell you __ but not __
analysis of variance; whether your continuous variable is different among more than two groups; whether a difference exists; which groups are significantly different from each other
what stats test is this an example of: is there a difference in hippocampal volume between rs11136000 genotypes C/C, C/T, and T/T?
ANOVA
correlation/linear regression: use to evaluate __
how associated two measurable variables are
what stats test is this an example of: is hippocampal volume smaller in adults?
correlation/ linear regression
when people say ‘regression’, they mean __
ordinary least-squares regression
linear regression: __ is the vertical distance from the line
the residual error
linear regression: the residual error is __
the vertical distance from the line
linear regression: __ is the beta / the effect of the association
the slope
linear regression: the slope is __
the beta / the effect of the association
regression line is __ and it __
the best fit for the data; minimizes the vertical distances between the data points and the line
linear regression graph: the dependent variable is on the __, the independent variable is on the __
y-axis; x-axis
Pearson’s correlation coefficient: aka __ or __
r-value or Pearson’s r
Pearson’s correlation coefficient: measures __, indicates __, ranges from __
how associated two variables are; direction of relationship; -1 to 1
a Pearson’s r of 0 means __
no correlation
a positive Pearson’s r
more of A is associated with more of B
a negative Pearson’s r
more of A is associated with less of B
r^2 = __ : measures __; does not tell you __
coefficient of determination; the proportion of the variance in the dependent variable that can be predicted by the independent variable; direction of the relationship
Pearson’s r: (absolute) values for no relationship, strong relationship, moderate relationship, weak relationship, perfect correlation
0, >.69, 0.30 - 0.69, <0.30, 1
r^2 ranges from __ to __
0 (no correlation); +1 (perfect correlation)
slide 32?
..
correlation/ linear regression: if the data are not __, they may be transformed by __ before performing the correlation
normally distributed; logarithm, square root, etc; ex: instead of using x in your analysis, use log10(x)
linear regression: note: it is really the __ that need to be normally distributed rather than __
residuals; the data itself
a __ test like __ should be used instead of a Pearson’s correlation if:
you have __ (3)
non-parametric; Spearman Rank correlation; small sample (<30); data outliers; non-normally distributed residuals that are not fixed through data
___ reduces the effect of outliers
Spearman Rank correlation
to perform a Spearman Rank, __ and assign __. values that are the same will __.
sort by each variable; a rank value to the values in order; share the rank halfway between their rank values
1 issue with interpretation of correlation/ linear regression
correlation does NOT = causation
ex: “We found that older adults who reported a greater consumption of fatty fish in the prior year had larger hippocampal volume. Our results suggest that fatty fish protects the brain from age-related atrophy.” what is wrong with the authors’ conclusions?
many other factors could influence hippocampal volume in a correlational study, and some of those may also be correlated with fatty fish consumption: higher socioeconomic class, general health consciousness, lower consumption of something else that is detrimental etc.
what is this an example of? does FA decrease with age in any voxel?
correlation/ linear regression: voxelwise brain imaging
correlation/linear regression: voxelwise brain imaging: each voxel in each brain is __
associated with a value (ex: DTI FA)
correlation/linear regression: voxelwise brain imaging: different brains are __; stats analyses compare __
transformed into a template space; corresponding voxels across subjects with a measure of interest
small sample should use more smoothing, decent sample should use less, huge doesnt need smoothing at all
..
multiple linear regression: similar to __; __
linear regression; other variables are added to the model to explain variability in the dependent variable that is not explained by the independent variable of interest
what stats test is this an example of? is hippocampal volume related to APOE4 genotype after removing the variability explained by age, sex, and education?
multiple linear regression
mult. linear reg.: global p tells you __, not __
whether the whole model is significantly associated with your dependent variable; whether your variable of interest is
when interpreting mult. linear reg. you must evaluate __ using __
contribution of individual independent variables; partial correlations
global p value is from __
multiple linear regression
mult. linear reg.: global p value must be significant before __
the contribution of any variable can be considered significant
when to use a non-parametric test such as Spearman Partial Correlation: __(3)
small samples, outliers, and residuals that are not normally distributed
a single multiple regression doesn’t test __, only __
mediation; how much one variable explains the variability in another
Spearman partial correlation tests are the __ equivalent of __
non-parametric; multiple regression
as with Spearman correlation tests, in spearman partial correlations the variables are __, and then __
ranked; multiple regression is run on the ranks
spearman partial correlation: to test mediations, you should use __ or __
pathway analyses; a specific series of regression analyses
spearman partial correlation: to test __, you should use pathway analyses or a specific series of regression analyses
mediations
covariate
a variable that may predict part of the variability of the dependent variable
a variable that may predict part of the variability of the dependent variable
covariate
types of covariates (2)
confounding variable, interacting variable
why use covariates? (2)
they help remove unexplained variability from the relationship that interests you possibly making that relationship clearer; they help ensure that the effect you see between the dependent variable and independent variable of interest isn’t actually due to something else that is correlated with both
ex of covariate: hippocampal volume in schizophrenia patients vs controls
antipsychotic meds may decrease hippocampal volume and only patients would be taking them
ex of covariate: hippocampal volume in schiz vs controls. why control for antipsychotic use in this study?
it’s correlated with diagnosis and may be correlated with hippocampal volume
ex of covariate: hippocampal volume in schiz vs controls. control for antipsychotics so that we know __
whether the measured effect between patients and controls is independent of or driven by the medication
You add a confounding covariate and the p value for the relationship between your dependent variable and your independent variable of interest becomes less significant. What could make that happen? (2)
the covariate is correlated with both the dependent variable and the independent variable of interest. the covariate is not correlated with either the dependent or the independent variable of interest (adding an extra variable slightly decreases statistical power)
You add a confounding covariate and the p value for the relationship between your dependent variable and your independent variable of interest becomes more significant. What could make that happen?
the covariate is correlated with the dependent variable, but is not very correlated with the independent variable of interest. controlling for it reduced unexplained variability in the dependent variable, allowing the relationship of interest to become clearer
a variable/covariate can still affect an outcome, even if __
there are no significant differences in that variable /covariate between groups
interaction variables: a variable whose value __
helps explain the relationship between another independent variable and the dependent variable
variable whose value helps explain the relationship between another independent variable and the dependent variable
interaction variable