SlideShare ist ein Scribd-Unternehmen logo
1 von 4
© TNS
Statistical Significance and Margin of Error
in TRI*M Studies
© TNS
Statistical Significance and Margin of Error
in TRI*M Studies
2
Generally, all standard rules for statistical significance also apply to TRI*M studies. In order to
determine the level of statistical margin the following parameters have to be taken into
account:
 Sample size
 Customized sample designs (e.g. stratified sampling of different customer segments)
 Size of universe
 Standard deviation
 Significance level
The mathematical calculation of the statistical margin of error for the TRI*M Index for various
sample sizes is based on:
 Normal distribution of TRI*M Index
 Infinite universe
 Significance level of 95%, 90% and 80%
 The challenge is to determine the effect of high quality sample designs. We therefore
calculate the standard deviation from all TRI*M Indices recorded in the TRI*M database. All
these scores are based on highly customized sample designs of our customer research
programs
© TNS
Statistical Significance and Margin of Error
in TRI*M Studies
3
The standard deviations for the following error rates for the TRI*M Index are calculated from
the TRI*M Database that includes more than 17,000 TRI*M Customer Retention studies with
more than 15 Mio. TRI*M Customer Retention interviews for 1,600+ companies.
 Most importantly, in order to apply these error rates to TRI*M Index comparisons over time
it is absolutely essential to keep the sample design stable and robust from wave to wave
 Any changes in the design impact the index and error rates!
Sample Size Confidence Coefficient
95% 90% 80%
n = 30 +/- 10.7 +/- 9.0 +/- 7.0
n = 50 +/- 8.3 +/- 7.0 +/- 5.4
n = 75 +/- 6.8 +/- 5.7 +/- 4.4
n = 100 +/- 5.9 +/- 4.9 +/- 3.8
n = 200 +/- 4.2 +/- 3.5 +/- 2.7
n = 300 +/- 3.4 +/- 2.8 +/- 2.2
n = 500 +/- 2.6 +/- 2.2 +/- 1.7
Example:
In case of a sample size of
n=200, the „true“ TRI*M
Index lies with a probability
of 90% within a range of
+/- 3.5 points
© TNS
Significant vs. meaningful differences
4
Statistical significance:
In statistics, a result is called statistically significant if it is unlikely to have occurred by chance.
Statistical significance heavily depends on the number of respondents in a survey.
Example:
 In a survey with n=8,000 respondents, a difference in the TRI*M Index (e.g. between two waves of the
survey) of 0.5 points can be considered as significant on a 90% significance level.
 This means that with a probability of 90% (or more), the TRI*M Index in the two groups (e.g. waves)
really is different (= not equal).
 But a significance does not give any information about the extent of this difference.
Meaningful differences:
A difference in results is called meaningful if it is of (business) relevance for the company. The
question, whether a difference in the TRI*M Index is meaningful or not, is completely independent of the
number
of respondents.
In the example, the difference in the TRI*M Index of 0.5 would be statistically significant (= not occurred by
chance), but the difference is small enough to be utterly unimportant. The company will not perceive any
difference in customer behaviour.
For the TRI*M Index, a difference of 3 points can be considered as meaningful
The use of the word significance in statistics is different from the standard one, which
suggests that something is important or meaningful

Weitere ähnliche Inhalte

Was ist angesagt?

Data array and frequency distribution
Data array and frequency distributionData array and frequency distribution
Data array and frequency distribution
raboz
 
Chapter 02
Chapter 02Chapter 02
Chapter 02
bmcfad01
 
Probability and statistics (frequency distributions)
Probability and statistics (frequency distributions)Probability and statistics (frequency distributions)
Probability and statistics (frequency distributions)
Don Bosco BSIT
 
2.3 Histogram/Frequency Polygon/Ogives
2.3 Histogram/Frequency Polygon/Ogives2.3 Histogram/Frequency Polygon/Ogives
2.3 Histogram/Frequency Polygon/Ogives
mlong24
 

Was ist angesagt? (20)

Descriptive statistics and anova analysis
Descriptive statistics and anova analysisDescriptive statistics and anova analysis
Descriptive statistics and anova analysis
 
Data Analysis
Data AnalysisData Analysis
Data Analysis
 
QT1 - 02 - Frequency Distribution
QT1 - 02 - Frequency DistributionQT1 - 02 - Frequency Distribution
QT1 - 02 - Frequency Distribution
 
Frequency Distribution
Frequency DistributionFrequency Distribution
Frequency Distribution
 
Data array and frequency distribution
Data array and frequency distributionData array and frequency distribution
Data array and frequency distribution
 
Chapter 02
Chapter 02Chapter 02
Chapter 02
 
Probability and statistics (frequency distributions)
Probability and statistics (frequency distributions)Probability and statistics (frequency distributions)
Probability and statistics (frequency distributions)
 
Boxplot
BoxplotBoxplot
Boxplot
 
Quartile Deviation
Quartile DeviationQuartile Deviation
Quartile Deviation
 
Statistics Based On Ncert X Class
Statistics Based On Ncert X ClassStatistics Based On Ncert X Class
Statistics Based On Ncert X Class
 
2.1 frequency distributions for organizing and summarizing data
2.1 frequency distributions for organizing and summarizing data2.1 frequency distributions for organizing and summarizing data
2.1 frequency distributions for organizing and summarizing data
 
3 goals calculate the arithmetic mean, weighted mean,
3 goals calculate the arithmetic mean, weighted mean,3 goals calculate the arithmetic mean, weighted mean,
3 goals calculate the arithmetic mean, weighted mean,
 
Hierarchical clustering and topology for psychometric validation
Hierarchical clustering and topology for psychometric validationHierarchical clustering and topology for psychometric validation
Hierarchical clustering and topology for psychometric validation
 
Statistics and probability lesson5
Statistics and probability lesson5Statistics and probability lesson5
Statistics and probability lesson5
 
Data Presentation using Descriptive Graphs.pptx
Data Presentation using Descriptive Graphs.pptxData Presentation using Descriptive Graphs.pptx
Data Presentation using Descriptive Graphs.pptx
 
Data Science Meetup: DGLARS and Homotopy LASSO for Regression Models
Data Science Meetup: DGLARS and Homotopy LASSO for Regression ModelsData Science Meetup: DGLARS and Homotopy LASSO for Regression Models
Data Science Meetup: DGLARS and Homotopy LASSO for Regression Models
 
Frequency distribution
Frequency distributionFrequency distribution
Frequency distribution
 
Frequency Distribution
Frequency DistributionFrequency Distribution
Frequency Distribution
 
Organizing-Qualitative-and-Quanti.-data
Organizing-Qualitative-and-Quanti.-dataOrganizing-Qualitative-and-Quanti.-data
Organizing-Qualitative-and-Quanti.-data
 
2.3 Histogram/Frequency Polygon/Ogives
2.3 Histogram/Frequency Polygon/Ogives2.3 Histogram/Frequency Polygon/Ogives
2.3 Histogram/Frequency Polygon/Ogives
 

Ähnlich wie Trim statistical significance and margin of error 2013

Statistical ProcessesCan descriptive statistical processes b.docx
Statistical ProcessesCan descriptive statistical processes b.docxStatistical ProcessesCan descriptive statistical processes b.docx
Statistical ProcessesCan descriptive statistical processes b.docx
darwinming1
 
Computing Descriptive Statistics © 2014 Argos.docx
 Computing Descriptive Statistics     © 2014 Argos.docx Computing Descriptive Statistics     © 2014 Argos.docx
Computing Descriptive Statistics © 2014 Argos.docx
aryan532920
 
Computing Descriptive Statistics © 2014 Argos.docx
Computing Descriptive Statistics     © 2014 Argos.docxComputing Descriptive Statistics     © 2014 Argos.docx
Computing Descriptive Statistics © 2014 Argos.docx
AASTHA76
 

Ähnlich wie Trim statistical significance and margin of error 2013 (20)

Lec 5 statistical intervals
Lec 5 statistical intervalsLec 5 statistical intervals
Lec 5 statistical intervals
 
5.DATA SUMMERISATION.ppt
5.DATA SUMMERISATION.ppt5.DATA SUMMERISATION.ppt
5.DATA SUMMERISATION.ppt
 
Hm306 week 4
Hm306 week 4Hm306 week 4
Hm306 week 4
 
Hm306 week 4
Hm306 week 4Hm306 week 4
Hm306 week 4
 
Measures of Central Tendency, Variability and Shapes
Measures of Central Tendency, Variability and ShapesMeasures of Central Tendency, Variability and Shapes
Measures of Central Tendency, Variability and Shapes
 
Lu2 introduction to statistics
Lu2 introduction to statisticsLu2 introduction to statistics
Lu2 introduction to statistics
 
4. six sigma descriptive statistics
4. six sigma descriptive statistics4. six sigma descriptive statistics
4. six sigma descriptive statistics
 
1.0 Descriptive statistics.pdf
1.0 Descriptive statistics.pdf1.0 Descriptive statistics.pdf
1.0 Descriptive statistics.pdf
 
Churn Modeling-For-Mobile-Telecommunications
Churn Modeling-For-Mobile-Telecommunications Churn Modeling-For-Mobile-Telecommunications
Churn Modeling-For-Mobile-Telecommunications
 
statistics.ppt
statistics.pptstatistics.ppt
statistics.ppt
 
Lecture-1.ppt
Lecture-1.pptLecture-1.ppt
Lecture-1.ppt
 
Lecture 1.ppt
Lecture 1.pptLecture 1.ppt
Lecture 1.ppt
 
Lecture 1.ppt
Lecture 1.pptLecture 1.ppt
Lecture 1.ppt
 
Statistical ProcessesCan descriptive statistical processes b.docx
Statistical ProcessesCan descriptive statistical processes b.docxStatistical ProcessesCan descriptive statistical processes b.docx
Statistical ProcessesCan descriptive statistical processes b.docx
 
Computing Descriptive Statistics © 2014 Argos.docx
 Computing Descriptive Statistics     © 2014 Argos.docx Computing Descriptive Statistics     © 2014 Argos.docx
Computing Descriptive Statistics © 2014 Argos.docx
 
Computing Descriptive Statistics © 2014 Argos.docx
Computing Descriptive Statistics     © 2014 Argos.docxComputing Descriptive Statistics     © 2014 Argos.docx
Computing Descriptive Statistics © 2014 Argos.docx
 
Estimation and confidence interval
Estimation and confidence intervalEstimation and confidence interval
Estimation and confidence interval
 
Bbs11 ppt ch08
Bbs11 ppt ch08Bbs11 ppt ch08
Bbs11 ppt ch08
 
Topic-1-Review-of-Basic-Statistics.pptx
Topic-1-Review-of-Basic-Statistics.pptxTopic-1-Review-of-Basic-Statistics.pptx
Topic-1-Review-of-Basic-Statistics.pptx
 
STATISTICAL PROCEDURES (Discriptive Statistics).pptx
STATISTICAL PROCEDURES (Discriptive Statistics).pptxSTATISTICAL PROCEDURES (Discriptive Statistics).pptx
STATISTICAL PROCEDURES (Discriptive Statistics).pptx
 

Trim statistical significance and margin of error 2013

  • 1. © TNS Statistical Significance and Margin of Error in TRI*M Studies
  • 2. © TNS Statistical Significance and Margin of Error in TRI*M Studies 2 Generally, all standard rules for statistical significance also apply to TRI*M studies. In order to determine the level of statistical margin the following parameters have to be taken into account:  Sample size  Customized sample designs (e.g. stratified sampling of different customer segments)  Size of universe  Standard deviation  Significance level The mathematical calculation of the statistical margin of error for the TRI*M Index for various sample sizes is based on:  Normal distribution of TRI*M Index  Infinite universe  Significance level of 95%, 90% and 80%  The challenge is to determine the effect of high quality sample designs. We therefore calculate the standard deviation from all TRI*M Indices recorded in the TRI*M database. All these scores are based on highly customized sample designs of our customer research programs
  • 3. © TNS Statistical Significance and Margin of Error in TRI*M Studies 3 The standard deviations for the following error rates for the TRI*M Index are calculated from the TRI*M Database that includes more than 17,000 TRI*M Customer Retention studies with more than 15 Mio. TRI*M Customer Retention interviews for 1,600+ companies.  Most importantly, in order to apply these error rates to TRI*M Index comparisons over time it is absolutely essential to keep the sample design stable and robust from wave to wave  Any changes in the design impact the index and error rates! Sample Size Confidence Coefficient 95% 90% 80% n = 30 +/- 10.7 +/- 9.0 +/- 7.0 n = 50 +/- 8.3 +/- 7.0 +/- 5.4 n = 75 +/- 6.8 +/- 5.7 +/- 4.4 n = 100 +/- 5.9 +/- 4.9 +/- 3.8 n = 200 +/- 4.2 +/- 3.5 +/- 2.7 n = 300 +/- 3.4 +/- 2.8 +/- 2.2 n = 500 +/- 2.6 +/- 2.2 +/- 1.7 Example: In case of a sample size of n=200, the „true“ TRI*M Index lies with a probability of 90% within a range of +/- 3.5 points
  • 4. © TNS Significant vs. meaningful differences 4 Statistical significance: In statistics, a result is called statistically significant if it is unlikely to have occurred by chance. Statistical significance heavily depends on the number of respondents in a survey. Example:  In a survey with n=8,000 respondents, a difference in the TRI*M Index (e.g. between two waves of the survey) of 0.5 points can be considered as significant on a 90% significance level.  This means that with a probability of 90% (or more), the TRI*M Index in the two groups (e.g. waves) really is different (= not equal).  But a significance does not give any information about the extent of this difference. Meaningful differences: A difference in results is called meaningful if it is of (business) relevance for the company. The question, whether a difference in the TRI*M Index is meaningful or not, is completely independent of the number of respondents. In the example, the difference in the TRI*M Index of 0.5 would be statistically significant (= not occurred by chance), but the difference is small enough to be utterly unimportant. The company will not perceive any difference in customer behaviour. For the TRI*M Index, a difference of 3 points can be considered as meaningful The use of the word significance in statistics is different from the standard one, which suggests that something is important or meaningful