Flashcards
General MCQ Statistics 1
Cards in this deck (62)
- 1why is the range not a great descriptive statistic?doesn't tell you how the measurements are distributed within the data set
- 2what is the sum of squares?all the values must be positive, so you square each residual (variance from the mean) and add them together
- 3then to average the amount of variation? what do you do? variance (ssquared)sum of squares/ degrees of freedom
DF = n - 1 - 4if you are given the variance, how do you then find the standard deviation? (s)square root the variance
- 5why does DF = n-1because the final data point entry is considered 'redundant'
- 6what is the major assumption when calculating the mean and standard deviation?that the data follows a normal distribution
- 7define outliersdata points a long way from the bulk of the data
- 8is the mean or median more sensitive to the presence of outliers?mean
- 9does the mean or median better express the data when there are outliers?median
- 10when there is a non-central/ bi-modal distribution where the mean value isn't actually a value, is the mean useful?no, misleading and unhelpful
- 11need for replication?cannot generalise findings from a sample of 1 --> need a group
the more times a result is observed, the less likely it will be due to chance
replication reduces the noise due to random variation - 12why do we use only samples of populations?realistic - can't measure whole pop.
take a representative sample and use to estimate population parameters of interest - 13why does haphazard sampling not equal random sampling?if you put hand in select, there may be reasons such as size etc that others may be picked earlier
- 14what is pseudo replication?you can't increase the n by taking multiple measures from the same individual
- 15what is a probability distribution?shows the theoretical distribution of all probabilities of all possible values of the random variable
total probability = 1 - 16give an example of discrete datacounts of patients arriving at doctor's surgery
can't have 1.5 patients - 17give an example of continuous dataheight of males in the population
- 18what does the area under the curve in a normal distribution = ?1
- 19what is a Poisson distribution?a discrete probability distribution that expresses the probability of a given number of events occurring in a fixed interval of time and/or space if these events occur with a known average rate and independently of the time since the last event
- 20—poisson distribution
the horizontal axis is the index k, the number of occurrences. λ is the expected value. The function is defined only at integer values of k. The connecting lines are only guides for the eye. - 21what is another word for normal distribution?gaussian distribution
- 22if the normal distribution of data is symmetrical about the mean then what does the mean also equal?median
- 23population standard deviation symbol?
- 24population mean symbol?
- 25what are z-scores?standard deviates
- 26how do you calculate a z -scorex - mean / SD
x = value that you are interested in - 27approximately 68% of the area under a normal distribution curve lies between what standard deviation values?+1 and -1
- 2895% of the curve area lies between what SD values?-1.96 and +1.96
- 2999%?+/- 2.58
- 3099.9%?+/- 3.29
- 31how can we assess sampling error?by taking repeated samples of size n
then plot a normal distribution all the sample means
if the underlying data is normally distributed then so should the sample means - 32equation for SEM?
- 33what is SEM?standard error of the mean
the standard deviation of the newly formed sample mean distribution - 34what happens to SEM as sample size increases?decreases
- 35the mean of the normally distributed sample will lie within +/- 1.96 (-2) standard deviations from the mean what proportion of the time?95%
the sample we take will most likely lie between +2 and -2 standard errors of the population mean - 36what are confidence intervals?they define boundaries within which we can confidently expect the population mean to fall
- 37to work out 95% confidence intervals for the mean what is the equation?sample mean +/- 2.00 x SEM
we can be 95% sure and gives you a range - 3899% CI? 99.9% CI?sample mean +/- 2.6 x SEM
sample mean +/- 3.3 x SEM - 39for small sample sizes, why should we use t-distribution rather than normal distribution when calculating confidence intervals?small sample sizes would mean than the data are not normally distributed
- 40does larger sample size lead to better confidence intervals?yes
- 41question - in our estimate of the mean age at first birth of 27.01 years in women with bleeding disorders, from our sample, different from mean of 25 years in the general population?we can say that it is different with a confidence of 99%
the range that 99% confidence intervals give doesn't overlap with 25 --> significantly different and there is no chance the mean could lie below 25
we can't be 99.9% confident because that range overlaps 25 --> could lie below 25 - 42what is the problem with using standard deviation bars?we can tell nothing about the significance of difference between means
only tells us how variable the population is - 43if the confidence interval bars on a bar graph don't overlap then what can we say?means are likely to differ with 95% confidence (or other stated confidence as longs the ranges don't overlap) p<0.05
- 44if the confidence intervals do overlap then what can we say?CANNOT say whether or not the two means are likely to differ significantly
--> carry out experiment 1000 times and measure the difference between the means on each of those 1000 occasions and get SEM differences between each group - 45what is the most common form of error bars on a mean?STANDARD ERROR
- 46if standard error DO NOT overlap then what can we conclude? if standard errors DO overlap then what can we say?means they may or may not differ at p< 0.05
means do not differ at p<0.05 - 47does a sufficiently small p value allow use to reject the null hypothesis?yes
Ho can be rejected if p < 0.05 - 48what does p < 0.05 mean?less than 5% probability of the results arising by chance (1 in 20) that the null hypothesis is true
- 49why is it said that hypothesis testing deals in probabilities and not in certainties?can never tell you whether anything is true or false
ONLY WHETHER THEY ARE LIKELY TO BE TRUE OR FALSE - 50what are the 2 ways to test hypotheses?two tailed - NO DIRECTION
one tailed - DIRECTION - 51what would the hypotheses look like for a two-tailed test?Ho = there is no difference between the means
H1 = there is a difference between the means
TESTS FOR DIFFERENCES WITHOUT SPECIFYING THE DIRECTION - 52what would the hypotheses look like in a one-tailed test?Ho = the mean in group B is = or < group A
H1 = the mean in group A > group B
TEST FOR > OR < HAS TO BE SPECIFIED IN THE HYPOTHESIS BEFORE THE DATA IS COLLECTED - 53in a two-tailed test, the 5% of 'non-confidence' is split how?2.5% either end, hence the name 'two-tailed'
one-tailed test has all 5 percent in only one of the tails - 54how do you calculate a one-tailed value from a two-tailed value?dividing by 2
- 55steps of carrying out a one-tailed t-test?1) t = (sample mean - hypothesised mean) / SEM
2) look up the corresponding p value in the t-distribution with n-1 degrees of freedom
3) then reject or accept the null hypothesis depending on the p value produced - 56what is the one-sample t test known as a PARAMETRIC test?these tests are performed on numeric variables and can be added and subtracted from a normally distributed sampling distribution
- 57what are the one-tailed t-test assumptions?1) must be normally distributed
2) must have large sample size so that normality can be assumed - 58what type of error is a type 1 (alpha) error?false positive
- 59a false negative is what type of error?type 2 (beta)
we will always make these errors so we try to limit them as much as possible - 60type 1 error rate is controlled by what?p < 0.05
- 61type 2 error rate is controlled by what?1 - power
if you accept a statistical power of 80% it will mean that you have a type 2 error rate (beta) of 20% (0.20)
if you conducted the experiment 20 times, 4 of the times (20%) if there was actually something going on, you would falsely conclude that there was nothing going on - 62why is a high statistical power of the experiment necessary?lowers the chance of a type 2 error being made
Information
- Description
- 2nd Year Statistics Flashcards on General MCQ Statistics 1, created by William Chapman on 11/05/2016.
- Created by
- William Chapman
- Views
- 24
- Rating