Sunteți pe pagina 1din 18

|   

   

 
 
  
Pradeep Gangadharan

Marketing Research Session 1


=      
      !
 uasic concepts pertaining to hypothesis testing
 Steps in conducting a hypothesis test
¢      !

  
 Specific hypothesis tests: (Parametric tests) z-test, t-test, chi-square test
 Examples of Univariate test situations
 "# !   $   
 ! 
 Spearman͛s rank correlation, Pearson correlation coefficient
[ %     
‰ Elements of a research proposal
 |    !

Marketing Research Session ¢


     !
     !      !
 &   '(
       '(
‰  ull hypothesis should be stated more conservatively; that is , failure to
reject null hypothesis from the sample evidence should preserve the
status quo
‰  ull hypothesis should always be specific hypothesis
‰ Generally the null hypothesis is something that we wish to reject, and
alternative hypothesis represents all other possibilities

 Î     åf 1 in 10 people are aware of the Medimix brand, we will


accept the null hypothesis that ͞Awareness activity for Medimix has yielded
positive results and we have to continue with the activity͟

Marketing Research Session 3


     !) 

 % *
% **  

Î  
  |  
H0 (true) Correct decision Type å error (ɲ error)
Confidence level (1-ɲ) Significance level
H0 (false) Type åå error (ɴ error) Correct decision
Power (1-ɴ)

Marketing Research Session 4


     !) 
¢
     !      !
  ! +     
 +
   
‰ The upper-bound probability of a Type å error refers to the significance
level (ɲ)
‰ The complement of Significance level is Confidence level (1 ʹ ɲ)
‰ The probability of rejecting the  ull hypothesis when it is false is
referred to as power (1- ɴ) ʹ probability of avoiding a Type åå error

 % , 

 , 
 
‰ Two-tailed tests reject the null hypothesis if, say, the sample mean is
significantly higher or lower than the hypothesised value of the mean of
the population (e.g.  ull hypothesis is pop. Mean = x and Alternative
hypothesis is Pop. Mean not equal to x)
(Acceptance and Rejection regions ʹ refer pg. 188, Kothari)

Marketing Research Session 5


     !) 

  
 !    
1. Set up H0 ( ull hypothesis) and Ha
¢. ådentify the nature of the sampling distribution curve and specific
appropriate test statistic (z-test / t-test)
3. Determine whether the hypothesis test is one-tailed or two-tailed
4. Taking into account the specified significance level, determine the critical
value (two critical values for two-tailed test) for the test statistic from the
appropriate statistical table
5. State the decision rule for rejecting H0
6. Compute the value for the test statistic from the sample data
7. Using the decision rule specified in Step 5, either reject H0 or reject Ha

Marketing Research Session 6


¢     !

  
-     !         

1. The number of variables to be analysed and
¢. The nature of data collected on each variable

%    +     

 Univariate analysis
‰ ås appropriate when just one variable is the focus of the analysis
 Multivariate analysis
‰ Two-tailed tests reject the null hypothesis if, say, the sample mean is significantly
higher or lower than the hypothesised value of the mean of the population (e.g.
 ull hypothesis is pop. Mean = x and Alternative hypothesis is Pop. Mean not
equal to x) (SHOW Acceptance and Rejection regions ʹ refer pg. 188)
¢%    +
  
   
  on-parametric procedures ʹ analysis techniques suitable for non-metric data
(nominal or ordinal data)
 Parametric procedures ʹ analysis techniques suitable for metric data (interval or ratio
data)
Marketing Research Session 7
¢     !

  ) 

$ ,.   !  /  +


 ås a means of formally checking the relationship between nominal / ordinal
(categorical) variables
 Will be able to tell whether there is a relationship or not ʹ Will not indicate
STRE GTH OF RELATåO SHåP!
 Computing the chi-square test statistic requires comparing the Observed
cell frequencies (cross tabulation / contingency table) and Expected cell
frequencies
 ås always one-tailed
 We reject the H0 if the computed chi-square value falls under the critical
region OR the computed chi-square value is greater than the critical value
 Refer pg. ¢40 and pg. ¢4¢ (Kothari) for examples

Marketing Research Session 8


¢     !

  ) 

0, /, (refer pg. 197, Kothari for formulae)
 Are parametric tests undertaken with metric data
šThe risk of committing a Type åå error is especially high in t-tests involving small sample sizes

$  ! '   1   


1  /  (
  /      2 % 
Single mean Large Z-test
Single mean Small T-test
Single proportion Large Z-test
Two means uoth samples are large Z-test
Two proportions uoth samples are large Z-test
Two means Either sample is small T-test
Two means Small / large ʹ but T-test
samples are dependent

Marketing Research Session 9


¢     !

  ) 

3       , "#  
% +      + *     .     
  
 !
   

ås there a significant relationship between


When the
customers͛ highest level of schooling
association between
Chi-square (education) (measured, say, as a four-
two categorical
contingency test category ordinal variable) and whether or
variables is to be
not they recommend X company to a
examined
friend? (measured as an ordinal variable)

To test hypotheses
ås the average waiting time for Heritage
that compare the
Test for a single Supermarket͛s customers at the store͛s
population mean of
mean checkouts significantly greater than 10
a variable to a
minutes?
prespecified value

Marketing Research Session 10


¢     !

  ) 

3       , "#  
% +      +   *     .  
   
 !
   

To test hypotheses that


ås the proportion of households
Test for a single compare the population
owning DVDs in Erode
proportion proportion of a variable to a
significantly less than 0.3
prespecified value

ås the average expenditure per


To test hypotheses that
household on purchase of frozen
compare the population
Test of two means chicken / meat significantly
mean of a variable for two
higher in Coimbatore than in
separate populations
Erode?

Marketing Research Session 11


¢     !

  ) 

3       , "#  
%  +       +   *    .   
  
 !
   

On the basis of data collected from a


To test hypotheses that compare panel of households before and after
Test of two means two population means of the the special ad campaign for Tide
when samples are same variable when the data for detergent, is the mean purchase
dependent the test are collected from the volume of Tides per household
same set of sample units significantly higher after the
campaign than before?

To test hypotheses that compare ås the proportion of two-income


Tests of two the population proportion of a households significantly lower in the
proportions variable for two separate southern states than in the northern
populations states?

Marketing Research Session 1¢


"# !   
      ++  '(
 ås a measure of association between two ordinal variables (e.g. two sets of
ranks)
 More powerful than chi-square ʹ involves measuring strength of relationship
 ås between +1 (perfect direct association) to -1 (perfect inverse association),
with a value of 0 signifying no association
 åmplicit assumption made is that the differences between the ranks on the
two variables can be meaningfully compared (so, similar ranking is to be
given to both variables)
 Significance can be tested with t-statistics

(Refer Pg. 469 in Parasu)


rs = 1 ʹ [ (6 x d¢i) / (n(n¢-1)) ]
t = rs я(n-¢) / (1-r¢s)

Marketing Research Session 13


"# !   
      ++  '(
 Example scenario:
ås there a significant relationship between motivation levels of salespeople
and the quality of their performance? (Assume the data on motivation and
quality of performance are in the form of ranks, say, 1 through ¢0, for ¢0
salespeople who were evaluated subjectively by their supervisor on each
variable) (pg. 470, Parasu)

Marketing Research Session 14


=          
1. A sample of 400 female consumers of a product had a mean age of
33 with a standard deviation of 5, and a sample of 500 male
consumers had a mean age of 3¢ with a standard deviation of 3. can
the mean ages of the female and male populations of consumers be
considered significantly different at a significance level of 0.05?

¢. MC company has been running an advertising campaign for one of


its brands for quite some time. åt wants to continue running the
campaign only if at least ¢0 percent of the target audience is now
aware of the brand; otherwise it wants to discontinue the campaign.
A just-completed survey of random sample of 500 customers in the
target audience showed that 90 were aware of the advertised brand.
Should MC company discontinue the ad campaign on the basis of
this evidence? Assume that if the company does decide to
discontinue the campaign, it wants to be 95 percent confident that it
is making the correct decision.

Marketing Research Session 15


=          
3. The sales manager of a firm administered a standard, multi0-
item job satisfaction scale to a sample of the firm͛s sales force.
The manager then correlated the satisfaction scores that the
ages of the salespeople in the sample. The Pearson correlation
coefficient between satisfaction and age turned out to be 0.08.
On the basis of this evidence, the sales manager came to the
following conclusions: ͞A salesperson͛s age has little to do with
his or her satisfaction. Furthermore, as salespeople get older,
they continue to have the same average levels of job
satisfaction.͟ do you agree or disagree with the sales
manager͛s conclusions? Explain your answer.

Marketing Research Session 16


|    
  !
 Show sample research proposal and work out costing

Marketing Research Session 17


 .  

Marketing Research Session 18

S-ar putea să vă placă și