Statistics Momento Mistaki

How to write the discrete uniform distribution in function form

P(X=x) = a for Inequality of domain

table for discrete uniform distribution [done]

done

Comment on the suitability

comparision of model’s prediction to actual value suggests good or not good

Refinement to a discrete uniform distribution for cloud cover

reason (varies with location and month) so non-uniform

Explainatory variable

independent variable

Response Variable

Dependent variable

How to state with a reason which is the explainatory variable

x will be the explainatory variable because it affects y

What words must you not use when talking about the explainatory variable in relation to the response variable

CAUSES or EXPLAINS no causation

How to Interpret a gradient of a regression equation

every UNIT rise in BLAH results in a (CHANGE) in blip of GRADIENT VALUE (CONTEXT!!!)

State assumptions that BLAH needs to make to use her binomial model

Binomial conditions hidden in CONTEXT

How to find the probability of the first success being after m trials from a binomial distribution of success of the successes in n trials

(probability of failure)^(n-1)(probability of success)

If two values from outside the IQR have moved to inside the IQR but above Q₁ which two values will change on a boxplot and why

Median and Q3 because there are more values below (VALUE OF Median) than below

Given a normal model, a sample of n and starting with X > a how do you find the probability that NONE of the sample will have X < b (b>a)

Find 1 - P( a < X < b)/P(X > a) and put to the power of n g idk why not 1-p then divide?

Given a normal model, sample where n have (X > a) and m have no data, what is the probability that none of the sample will have (X < b)

(1 - P(a<X<b)/P(x>a))to the power n x P(X>b)to the power m

Use the value calculated from the data and the value predicted by the model to comment on the suitability of the model

compare values and if not roughly same say ‘so model may not be suitable’

explain why a linear regression model based on these data is unreliable when (a value is out of the range)

It requires extrapolation so will be unreliable

state what is measured by the product moment correlation coefficient

Linear association between x and y

what does mean temperature of 27.2 suggest in july

Overseas not perth

what does windspeed mean of 3.5 suggest

not coastal

Accuracy for probabilities (awrt)

3 s.f.

when hypothesis testing for mean from sample mean what do you use for σ

sd/√n

How to hypothesis test for mean

assume h0, write distribution of xbar with sd/rootn , find p(xabr is more extreme than value) compare probability with sig level

What accuracy should you use for the p you get from a hypothesis test when comparing with significance level? awrt

3 s.f.

how to show that events are NOT independent

P(A)xP(B) = a DOES NOT EQUAL P(A intersection B) = b

What accuracy for p found for a binomial distribution?

anything?

She believes that a non linear model is better, the pmcc of the coded values (THIS VALUE) is greater than that of the uncoded. Explain how this value supports her belief

The value is close(r) to 1 or there is strong(er) (positive)

correlation

When using answers to comment on model can you just use the question part labels? (i think?)

yes

Should you calculate the standard deviation in the little box in distribution mode

NO!

Describe the correlation (from a graph)

negative or positive

what is (apparently) not needed to “describe the correlation”

strength

weak negative correlation between Y and temperature for Perth with only 1 value at zero and one near zero for Y. no scale on Y. Suggest which variavle is on the y-axis

Rainfall or pressure

How to find the range from a box plot

highest value plotted - lowest value plotted (INCLUDE OUTLIERS!)

Correlation between windspeed and temperature in winter? guessing from MS comment

Might expect weakly positive?

Taruni defines an outlier as a value more than 3sds above the mean. State how many outliers T would say there are in these data, giving a reason for your answer.

mean + 2sd = a so only (however many values above a) outliers

Without carrying out any further calculations, explain w hy the standard deviation of all x times [ with added ones near the mean] will be lower than your answer to d [the standard deviatiojn without these two added ones]

Both values will be less than 1 standard deviation from the mean and so the

standard deviation of all x values will be smaller

when does standard deviation decrease when you add values?

when the values are within one standard deviation of the mean

what does standard deviation represent

average distance from the mean

find P( X < a | X > 0)

P(0<X<a)/P(X>0)

[the probability that the time, T, is less than 2 given its greater than zero greater than the probability that T is less than 2] hence explain why this normal distribution may not be a good model for T

The current model suggests non-negligible/significant probability of T values < 0 which is impossible

If you have a normal distribution and you think that in fact X must be greater than a how do you find the median of X?

Find the area that is greater than a, A. Find the value x such that the area greater than x is equal to half of A.

How to find the mean of coded data when something is added or subtracted to code x into y

mean of y and then uncode as if for x (show this in an equation)

How to find the standard deviation of data coded with addition or subtraction

standard deviation is not affected by this

type of coding (quote this)

{using your knowledge of the large data set and wind is blows clockwise if high pressure and anticlockwise if low and pressure is around 1029. put these wind directions with heathrow hurn and leuchars. } Give a reason for your answer

Wind direction is direction wind blows from

high pressure so clockwise

locations are (from north to sound) Leuchars, heathrow, hurn

so: answer

How do they seem to get method marks for inverse normals?

Standardising.

How to find the probability that an individual that has the feature x such that ( a < x < b ) is greater than q

P( q < x < b)/ P( a < x < b )

If you have probabilities in terms of loga, logb etc and none of the probabilites are zero what does it imply? and why

a > 1, b >1 etc because log1 = 0

what is the significance level?

probability of incorectly rejecting the null hypothesis

When hypothesis testing what is the sample used for testing for a proportion in the binomial distribution

used to make an

inference about the

population

what are correelation coefficinets broadly

measures of

how close data points lie to

a straight line

for 2 tailed tests what is the expected value of the binomial distrbution

np

what can difference smaples lead to

lead to different

conclusions about the

population