Question | Answer |
Discrete | Finite number |
Continuous | Infinite without gaps |
Nominal | Categories
No order |
Ordinal | Ordered but difference between is meaningless
Relative comparison
Ex:grades |
Interval | Ordered but meaningful
Does not start at 0
Ex:temp |
Ratio | Ordered/meaningful
Starts at 0
Ex:distance |
Systematic sample | Start point
Select every Kth element |
Convenience sample | Easy to collect
Close to researchers location |
Stratified sample | Subgroups (2 of them)
Same characteristics
Draw same amt of sample from each
Consistent |
Cluster sample | Divide into groups
Select only some groups
Choose all elements from selected group
Faster and less expensive |
Multistage sample | Combination of methods
Select sample in each stage
Each stage different method
Natural clusters |
Reason for Frequency Tables | 1)summarize large data
2)analyze nature
3)basis for graphs |
Relative Frequency | (Class frequency/sum of all frequency)x 100% |
Unusual Center | Mean-2s or mean+2s |
Coefficient of variation | (s/x)*100% |
Bimodal | 2 modes |
Multimodal | 2+ modes |
Midrange | (max # - min#)/2 |
Percentile | L(position)= k(percent)/100 *n |
Interquartile | Q3-Q1 |
5 Number Summary(mon,q1,med,q3,max) | 1-Vars Stats |
Odds against | P(not A)/P(A)-> A:B |
Payoff odds | Net profit: amount of bet |
Addition Rule of Probability | P(A or B)= P(A) + P(B) /total outcomes |
Complements | P(none)= 1 - P(at least one)
P(at least one) = P(only A or only B or both)
= P(A only) + P(B only) - P(both) |
Multiplication Rule (Independent) | P(A and B) = P(A) * P(B) |
Multiplication Rule (Dependent) | P(A and B) = P(A) * P(B/A)
after event A has occurred |
Conditional Probability (Independent) | P(B/A) = P(A and B) / P(A) |
Bayes' Theorem | P(A)*P(B/A) / [P(A)*P(B/A)]+[P(no A)*P(B/no A)] |
Fundamental Counting Rule | m*n ways
P(A) = 1/m*n |
Permutation(different) | nPr
n = # of items
r = amt selected |
Permutation(identical) | nPr = n!/ n1!n2!nk! |
Combination Rule | nCr
no repeats
no order |
Requirements for Probability Distribution | 1)Sum of P(x) = 1
2)0 < P(x) < 1 for every x |
Probability mean | Sum of x * P(x) |
Probability standard deviation | Square root of Sum of [(x^2 * P(x0]- mean^2 |
Unusual Probability | P(x or more) < 0.05
P(x or fewer)< 0.05 |
Binomial Distribution Requirements | 1)fixed # of trials
2)independent
3)2 categories
4)P(success)same in all success |
Binomial Probability | binompdf(n,p(success),x success) |
Binomial Mean | n*p(success) |
Binomial Standard Deviation | sq. root of n*P(success)*P(failure) |
Unusual values for Binomial | mean - 2stan. dev
mean + 2stan. dev |
Poisson Distribution Requirements | 1)x = event of interval
2)random
3)independent
4)uniformly distributed |
Poisson Standard Deviaton | sq. root of mean |
Using Poisson as Binomial Distribution
Requirements: | 1)n > 100
2) np < 100 |
Poisson Probability | poissonpdf(mean,x selected) |
Normal Distribution Characteristics | 1)bell-shape
2)mean = 0
3) s = 1 |
Uniform Distribution Characteristics | 1)area = 1
2) correspondence between area and prob |
Normal Distribution (Area under graph) | normalcdf(left z,right z) |
Z-score(normal distribution) | invNorm(area left of z-score) |
Finding P(individual value) w/ Norm Distr | z = x - mean/ stan. dev |
Finding P(sample) w/ Norm Distr | z = x - mean/ (stan dev./sq. root of n) |
mean & x to z-score | normalcdf(left, right, mean, s)= P(x)
invNorm(P(x))= z-score |
Find x value of nonstandard norm distr | invNorm(area to left, mean, s) |
Sample Variance | Sum(x - mean)^2 / n -1
mean = sum of x/n |
Normal Distr As Binomial Approximate
Requirements | 1)independent simple random sample
2)np> 5 and nq>5 |
Continuity Correction | x - 0.5 to x + 0.5 |
P(Area to left) | normalcdf(-99999,z-score) |
Normal Distribution on Graph | 1)straight line
2)no systematic pattern |
Find Critical Value(z-score) | 1-(confidence interval/2)-->invNorm(1-alpha) |
Margin of Error (z-score) | z(alpha/2) * sq. root(p(success)*q(failure)/n)
or
upper CI - lower CI/ 2 |
Confidence Interval(z- score) | Stat--> 1-PropZInt |
P(Success) | upper CI + lower CI/ 2 |
Sample Size(independent) | n = (z(alpha/2)^2 *p*q)/E^2
or n = (z(alpha/2)^2 *p*q*.25)/E^2 |
Margin of Error(stan. dev known) | ZInterval
Set mean = 0 |
Sample Size(stan dev known) | n= [z(alpha/2)*stan.dev/E]^2 |
t statistic | invT(1-alpha,df=n-1) |
Confidence interval(t-score) | TInterval |
T-Score Properties | 1)norm distr
2)different t for different n
3) mean = 0
4)stan. dev. > 1
5)n increase, t --> norm distr |
Chi-Square Distribution Properties | 1)not symmetric
2)positive values
3)different for each df |
Finding Chi-Square | 1)calculate alpha and df
2)What kind of test?
3)Look at Tables given |
Confidence interval(chi-square) | sq.root[(n-1)s^2/chi right] < stan. dev < sq. root[(n-1)s^2/chi left] |
Reject null (p-value) | p-value < alpha |
Find p-value | normalcdf(left, right) |
Reject null (test statistics) | test statistic falls in critical region bounded by critical value |
Reject null(confidence interval) | Confidence interval does not contain claimed value |
Type 1 Error | null true --> reject (alpha) |
Type 2 Error | null false --> fail to reject (beta) |
Two Tail Test | null =
alternate not = |
Right Tail Test | null =
alternate > |
Left Tail Test | null =
alternate < |
Testing Claim of Proportion | 1-PropZTest |
Test Claim on Mean (stan. dev known) | Z-Test |
Test Claim on Mean (stan dev Unknown) | T-Test |
Test on 2 Proportions | 2-PropZTest |
Confidence Interval(2 Proportions) | 2- PropZInt |
Test on 2 Means (stan. dev. unknown/ independent) | 2-SampTTest
2-SampTInt |
Test on 2 Means(Stan dev known) | 2-SampZTest
2-SampZInt |
Claim on Mean (Dependent/differences) | 1) L1 - L2
2)TTest
3)TInt |
Compare Variation of 2 Samples | 2-SampFTest |
Correlation | 1)Straight-line
2) r = LinRegTTest
3) r > critical value of alpha
4) reject null |
Regression | LinRegTTest |
marginal change | slope of regression line |
Residual | observed y - predicted y
y from table - y from regression line |
Coefficient of determination | r^2 |