POL 051 Final

controlled, randomized, assigned; can specifically see all changes which creates a very clear causal argument

Less control than experimental studies, however provide better data (not as clear causal argument)

as our sample size increases, our confidence in the representativeness of that sample increases

measures the dispersion of sample means around the population mean; calculated by taking the standard deviation of the samples and dividing it by the square root of the sample size

We recognize that there is potential for a badly drawn sample that may lead to wrong conclusions; concern for bias

each thing leads to the next; the argument is always moving forward

an argument runs in 1 direction, so what is causing what?

something is simply associated with another thing

converting a variable from a concept to something that defensibly measures that concept (how do you measure a variable)

look like whole numbers; a measurement with whole number values and unequal unit pair differences

data that exists that doesn't have a number assigned to it, data is described and we can rank them; no numerical order

take on a spectrum, can have decimals; any possible observable value in their range; equal pair unit differences

without variation in the IV, we have no explanatory power

without variation in the DV, we have nothing to study

tells use we are confident that a relationship between these 2 variables (based on p values)

the likelihood that you're wrong and what you found isn't true; the probability that your observed value is a product of randomness; smaller values indicate something is statistically significant (e.g. p-value of 0.01 = 99% chance you're right, 1% chance you're wrong)

tells you how confident you are that a true relationship lies between your confidence interval; helps us identify if there is a 0 in the confidence interval, if there is a 0 then the relationship is not statistically significant

if we took an infinite number of samples, their sampling distribution would be normally distributed around the population mean

ranges from -1 to 1, where smaller values indicate weaker correlations and larger values indicate a stronger correlation; negative numbers indicate negative relationships, and positive numbers indicate positive relationships (e.g. 00.19 very weak, 0.8-1.0 very strong)

provide one of the best ways of analyzing intersections of sortal/categorical data (easy to see, but sometimes hard to know its real or random intersections

test statistic that when combined with degrees of freedom will allow us to interpret the relationship; comparing the critical value to our chi test statistic, we can evaluate the statistical significance of the observed relationship

when you believe an independent variable that is categorical will have an effect on a scalar dependent variable; the more significant the difference, the bigger the t will be

when the t-statistic is greater than the critical value, we can conclude that the null hypothesis is incorrect, and a relationship does exist

1. Coefficient
2. Standard Deviation
3. P-Value
4. Confidence Interva

the independent effect of a one unit increase in the IV on the DV

measures essentially how well your estimated line fits the dependent variable

calculating the average distance between the actual data points and the line (smaller RMSE = better fitting line)

represents the proportion of the variation in the dependent variable that is accounted for by the regression; low R^2 is not explaining much of the variation, but high R^2 is explaining a lot of the variation

best indicator describing data that is ranked, but not significantly skewed

best indicator for data that is ranked and may be skewed

best indicator for when data isn't ranked or ordered

the difference between the highest and lowest scores in a distribution

measure of statistical dispersion, being equal to the difference between the upper and lower quartiles, IQR = Q3 − Q1.

broadly tells us how many observations fall within a certain distance of the mean; characterizes the spread of the distribution

the more spread out the distribution will be

the less spread out the distribution will be

When both variables are sortal

When both variables are scalar

Scalar/Continuous DV and Sortal IV

Sortal DV and Scalar/Continuous IV

assessed through either the p-value, t-statistic, or confidence intervals (tells you whether a result is statistically different from 0); requires you to interpret those results to say whether they matter and how much

when the mean is way larger or way smaller than the median (large and small outliers)

in a normal model, about 68% of values fall within 1 standard deviation of the mean, about 95% fall within 2 standard deviations of the mean, and about 99.7% fall within 3 standard deviations of the mean