Variance and Sampling Flashcards
When do we use standard error?
used when the amount of standardised deviation between the population and the sample needs to be estimated
(larger sample size = lower SE)
Why use stats?
To generalise from a sample we must use stats to assess the probability any effects will also be found in the population
What is the F ratio?
Effect over Error
Define Effect
Define error
effect = difference between groups/ conditions error = amount of difference expected between population and groups (without diff conditions affecting)
How is variance calculated?
calculated by determining how much each score differs from the mean average
square each value (SUM OF SQUARES)
add them up
divide by no. of scores (MEAN SQUARE)
(dividing by N = variance in the sample)
(dividing by N-1 = estimate of variance in population)
we can then square root in order to get back to where we started
(Square root of variance = Standard Deviation)
Sampling Distribution
when you plot sample stats from all of your samples as a frequency histogram
What would we use the SD of the sampling distribution to look at?
How would we find SD?
we need the SD of the sampling distribution to calculate the amount of deviation we would expect to see between a sample and the population
- if we plot the distribution of the samples we get ‘sampling distribution of the mean’, we expect this to be normally distributed
- To find the SD of sampling distribution just calculate SD (square root of variance) = this represents the amount of deviation between mean and each of our samples (can use this to compare to the effect we found)
Standard error definition?
its a measure of the degree to which the sample means deviate from the mean of the sample means
standard error of the mean = Sample SD divided by square root of n
Error bars show
different types? What they show?
they give us an idea of how much error there might be in our experiment. AND where the mean might lie in the population
- they contain 95% confidence intervals (upper and lower boundary)
SE and CI error bars based on population
SD error bars based on sample (deviation between sample data points and sample mean)
(SD error bar is least representative of t test as it is based on sample)
How would we use 95 % confidence intervals?
our sample mean is an estimate of the population mean - we don’t know if it is an underestimation or overestimation, so, we use confidence intervals
-the 95% confidence intervals takes into account SE and the fact our data should be normally distributed (in normal distribution 95% of the data falls up to 2SDs away from the mean)
mean +/- (2SDs)x (SE to the mean)
- illustrate how much deviation there is between our sample mean and population mean
- 95% CI indicates how much overlap we might have between our 2 groups in the population(large overlap = less likely you will find an effect in the sample)
Sampling
-what makes a good sample?
Sample must be normally distributed
Bias in sampling can be reduced by randomly assigning p’s to groups/conditions
Size of the sample must be sufficient enough to result in normal distribution(larger the better)
T-tests
T= mean 1 - mean 2 (mean diff) divided by SE of the differences
if the value is less than 1 your ERROR is LARGER than the effect (ie. the amount of deviation thats expected between sample and population mean is larger than the effect found)
