Vous êtes sur la page 1sur 16

N current N

3 9 306.194 This calculates the required N for 99% confidence level based on current ratings
3 9
3 9 176.714 This calculates the required N for 95% confidence level based on current ratings
3 9
3 9 125.235 This calculates the required N for 90% confidence level based on current ratings
4 16
4 16
4 16
4 16 0.737 This calculates the z-score based on the current sample.
4 16 To look up the confidence level, you need to use a normal curve distribution.
5 25 Excel can do this for you, and it will be more precise than manually looking it up.
5 25
5 25 0.770 This calculates the confidence level based on the z-score.
5 25 This is ultimately what you would want to report as the 'current' confidence level based
5 25
5 25
1 1 *Note: These values will all update as you change the numbers in column A.
1 1 To add more values, you should insert rows at the end of the ratings in the
1 1 columns (before the sums). Again, the formulas will update accordingly.
2 4
2 4
2 4
2 4
2 4
2 4

80 302 Sums of columns

25 N current
d on current ratings

d on current ratings

d on current ratings

lly looking it up.

rent' confidence level based on the sample obtained.


Rating 1 Rating 2 Difference in ratingsDifference squared
X1 X2 (X2-X1) X^2 Std. Dev = ((X^2 - (X)^2/N)/N-1)
1 3 2 4
2 3 1 1
3 2 -1 1 Std. Dev 1.761428846
4 4 0 0
5 5 0 0 t = (mean - X)/SEM
1 5 4 16 so t = mean difference/SEM
2 4 2 4
3 2 -1 1 SEM = sd/N
4 3 -1 1 SEM 0.393867464
5 5 0 0
1 2 1 1 t -13.83714194
2 5 3 9
3 3 0 0 df = N-1 19
4 3 -1 1
5 4 -1 1 p 0.000
1 5 4 16
3 5 2 4
4 4 0 0
5 4 -1 1 Sums of columns
5 3 -2 4 Mean difference
N
3.15 3.7 Means of columns 11 65
0.55
20
X^2 - (X)^2/N)/N-1) Written standard deviation formula

Operational std. dev. formula

X = 0 (assumes no difference between first and second rating)


ean difference/SEM Written t-statistic formula

Written standard error of mean formula


Operational SEM formula

Operational t formula

Operational df

Significance level p value is operational


Rule of thumb: if p < .05, we say it is a statistically significant difference.
That is, the two ratings were significantly different. To see which was higher, check the means from the first
two columns.

This is how you will report whether there is a significant difference in ratings to the client.

Again, all of these values are updating. If you want to add more scores, you'll want to insert cells from
Column A to Column G. This way the calculations will stay intact.
Proportions Test

H Null: There is no difference in participant responses (Yes, No)


H Alternative: There is a difference in participant responses

N >= 16 minimum sample size

Percent of participants indicating Yes 0.800 Numerator


Critical Vlaue % 0.500 Denominator
Total number of Responses 11 Results (Z)

If Z < -1.96 or greater than 1.96 =


This is the proportions test formula. We can use it to determine success of a
vote on any particular flag where response = "Do you agree with this: Yes, No."
We will set our critical value at .75 meaning that 75% of responses must agree
to the flag criteria before action will be taken.

0.300 The test will determine if we have met the 75% threshold with a margin of
0.151 error at 5%.
1.990 If we Accept the null Hypothesis then we are making the decision that a
majority of the community agrees with the Flag criteria.
Accept Null, p > .05
success of a
this: Yes, No."
es must agree

margin of
One rating Test

H Null: The mean rating does not differ from the midpoint rating
H Alternative: The mean rating does differ from midpoint rating (significantly higher or lower)

Rating scale 1 to 10
If normal distribution across scores, median/mean = 5.5*
*(Rating scale from 1 to 6, midpoint would be 3.5, etc)

Ratings Ratings
1 1 Midpoint (test-value): 5.5
3 9
2 4 Std. Dev = ((X^2 - (X)^2/N)/N-1)
5 25
7 49 Std. Dev 2.669116522
9 81
6 36 t = (X - test-value)/SEM
5 25
3 9 SEM = sd/N
1 1 SEM 0.629116797
2 4
10 100 t -2.031066064
2 4
2 4 df = N-1 17
4 16
3 9 p 0.058187527
7 49
4 16

Mean 4.222222
her or lower)

Written standard deviation formula

Operational std. dev. formula

Written t-statistic formula

Written standard error of mean formula


Operational SEM formula

Operational t formula

Operational df

Significance level
Trend Test 45
X Y XY X Y 40
Days Count (ratings)
1 5 5 1 25 35
2 8 16 4 64 30
3 12 36 9 144
4 15 60 16 225 25
5 20 100 25 400 20
6 25 150 36 625
7 28 196 49 784 15
8 32 256 64 1024 10
9 35 315 81 1225
10 38 380 100 1444 5
11 40 440 121 1600 0
12 41 492 144 1681 0 5 10
13 42 546 169 1764
14 42 588 196 1764
15 42 630 225 1764

Means 8 28.33333 What will be most useful for comparing rating


Std Devs* 4.320494 12.88755 value from a regression equation.
N (pairs) 15
b is equal to the regression coefficient multipli
r 0.969818 of the y variable divided by the standard devia

*Slightly different computation than used in ratings tests: This represents the slope of ratings over time.
((X^2 - (X)^2/N)/N) Written standard deviation formula highest b value therefore represents the item th
ratings over time, and thus is "trending"
sr = ((1 - r)/(N - 2)) Written standard error of r formula
For example, Item Y (left) had more ratings th
t = (r - )/sr Written t-statistic formula higher b value. The correlation values don't ho
(as you can see, the correlation r for Item Z is
test value = ("rho") = 0 less ratings)

0.067627 Operational std error of r formula

14.34076 Operational t formula

df = N-2 13 Operational df

2.4E-09 Significance level

b = r(sy/sx) 2.892857 Operational b formula


X Z XZ
Days Count (ratings)
1 5 5
2 8 16
3 10 30
Y Count (ratings) 4 12 48
5 13 65
Z Count (ratings) 6 14 84
7 15 105
8 18 144
9 20 180
10 22 220
11 23 253
10 15 20 12 25 300
13 27 351
14 28 392
15 30 450

seful for comparing rating trends will be the b or Beta Means 8 18


ion equation. Std Devs* 4.320494 7.474401
N (pairs) 15
ression coefficient multiplied by the standard deviation
ided by the standard deviation of the x variable. r 0.997117

slope of ratings over time. The item being rated with the *Slightly different computation than used in ratings tests
efore represents the item that is garnering the most ((X^2 - (X)^2/N)/N) Written standard deviation
nd thus is "trending"
sr = ((1 - r)/(N - 2)) Written standard error of r
Y (left) had more ratings than Item Z (right) and has a
correlation values don't hold value for assessing trend t = (r - )/sr Written t-statistic formula
correlation r for Item Z is higher than for Item Y despite
test value = ("rho") = 0

0.021045 Operational std error of r

47.37957 Operational t formula

df = N-2 13 Operational df

6.02E-16 Significance level

b = r(sy/sx) 1.725 Operational b formula


X Z

1 25
4 64
9 100
16 144
25 169
36 196
49 225
64 324
81 400
100 484
121 529
144 625
169 729
196 784
225 900

on than used in ratings tests:


Written standard deviation formula

Written standard error of r formula

Written t-statistic formula

Operational std error of r formula

Operational t formula

Operational df

Significance level

Operational b formula
Determining Current Confidence Level that Mean will Remain
within Margin of Error
To calculate current confidence level:
1. Input the values for X , X2, and Ncurrent described previously into the following equation

Order of Operations:
a. Calculate the denominator using steps a e from above.

b. Calculate the numerator by multiplying the square root of Ncurrent by 0.2

c. Divide the value from b. by the value of a. The result is the z-score associated with our confi

d. Look up confidence level using a z-table (aka normal distribution table) that shows cumulativ

*This table shows cumulative probabilities for z-scores: http://www.stat.wisc.edu/~ifischer/Statisti

*We wont ever get a negative z-score from the above equation so we will always have a minimum
the following equation:

ociated with our confidence level.

) that shows cumulative probabilities for z. Identify the z value from part c., then find the cumulative probabi

c.edu/~ifischer/Statistical_Tables/Z-distribution.pdf

ways have a minimum confidence level of 50%


he cumulative probability for that z. Convert the probability to a percent ( 100), which gives us the confiden
h gives us the confidence level as a percent.

Vous aimerez peut-être aussi