Editing the Data

The given responses are from my survey, "Are You Selfie Ready?" This is a screen shot from my drive.

To prepare these responses for analysis, we need to download it as excel file. Go to File, Download as and Microsoft Excel (.xlxs)

Select the location in your computer to save the file.

Open the excel file.

The responses are of two types, from those who are selfie types and from those who are not. We need to merge those responses.

Go to the cell from where it is to be super-imposed and select paste special.

From the Paste Special check box, Check the box for skip blanks.

The responses will be merged.

Delete the rest of the columns. The excel sheet is ready for coding.

Extended Multi-attribute Attitude Model

An Example of the Extended Attitude Model (The-Theory-of-Reasoned-Action)

The Extended Model is:

Let’s assume we are interested in evaluating consumer’s attitudes about laptops.  We determine that the relevant factors that influence purchase are Provides Good Support, Screen Size, Weight, Battery Life, and Price.

Keeping in mind the rationale behind the Extended Model, we ask the consumer to think about purchasing a laptop. The attitude is calculated using following steps.

Step 1: Establishing the E_I scores:

Consumers are asked to evaluate each of the following consequences of buying a laptop on a scale of 1 to 10.

Step 2: Determining the B_O scores on a scale of 1 to 10.

Buying the _____________ laptop will:

Step 3: Calculating A_IO

Step 4: Determining Social Influence on Purchase

Next, we want to determine the social influences on the purchase.  We would need to know who the consumer would likely be influenced by in order to ask these questions. 

Use this scale to indicate how each person feels about you taking each Laptop:

Buying the following brand of laptop:

The above scores are the NBs (the Normative Beliefs)

Now, we see how “motivated to comply” they are with each person……

How much do you want to do what each person thinks you should do?

Step 5: Putting weights

Finally, we need to obtain an indication of how much weight to put on each factor.  In a research study, these would be “beta weights” from a regression analysis.  For this example, we ask the following question:

When people purchase laptop, there are two important influences on their decision about what laptop to buy.  What they personal think about each laptop, and what people who are important to them think.  Please divide 100 points between these two factors to indicate their importance in your decision about what trip to buy (e.g., 80-20, 35-65, 50-50, etc.)

    My own beliefs _____      Other peoples' beliefs _____

For this example, let’s assume that person responds .8 for their own beliefs and .2 for other people’s.

Step 6: Calculating Behavioural Intentions

Step 7: Calculating B.I. Scores

Mathematical Model of Multi-Attribute Attitude Model

Let us identify the attitude for Online Travel Portals. The consumers are asked to evaluate the following attributes for online travel portals for holiday planning; holiday options, fares and taxes, website, hidden costs, customer communications, local support, payment options, working flexibility and product delivery and grievance handling

Let us rate these attributes on a scale of 1 to 10 for evaluation of the attribute being good or bad i.e. E_I 

 My ratings are:

The next task is to rate three brands of interest to determine whether the consumer believes each brand possesses each attribute (the B_O).

 To calculate the attitude about each brand of Online Travel Portal, multiply the attribute evaluations time the brand’s rating and sum for each brand:

I got the following Attitude scores:

Makemytrip = 681

Cox & Kings = 658               

Kesari = 626

what about you?

Next: Extended Multi-Attribute Model

Mutli-Attribute Attitude Models

Multi-attribute attitude models portray consumers’ attitudes with regard to an attitude object as a function of consumers’ perception and assessment of the key attributes or beliefs held with regard to the particular attitude object. There are many variations of the attitude model, three to consider are: attitude-toward-object model, attitude-toward-behavior model, and the theory-of-reasoned-action model.

Attitude-towards-object Model

This model is especially suited for measuring attitudes toward a product category or specific brands. According to this model, the consumer’s attitude toward a product or specific brands of a product is a function of the presence (or absence) and evaluation of certain product-specific beliefs and/or attributes. Consumers generally have a favorable attitude toward those brands they believe have an adequate level of attributes that they evaluate as positive, and they have unfavorable attitudes toward those brands they feel do not have an adequate level of desired attributes or have too many negative or undesired attributes. 

This can be explained in the car purchase example, where an individual is evaluating features of petrol and diesel cars.. The features can be evaluated into two categories as favourable and unfavourable.

The Attitude-Toward-Behavior Model

This model is designed to capture the individual’s attitude toward behaving or acting with respect to an object, rather than the attitude toward the object itself.

The appeal of this model is that it seems to correspond somewhat more closely to actual behavior than does the attitude-toward-object model. 

We have already discussed the relationship between attitude and behaviour. Having a positive or favourable attitude does not always result in buying behaviour. For example, an individual has a positive attitude towards BMW, but may not purchase a BMW because of financial constraints.

Attitude-towards-Behaviour Model aims at judging the relative importance of the features as per the benefits it offers as consumers majorly decide on the basis of functional benefits. 

Theory-of-Reasoned-Action Model

This model represents a comprehensive integration of attitude components into a structure that is designed to help knowing what and why of behaviour. It gives better explanation and predictions of behavior. The theory-of-reasoned-action aims at analyzing intentions vis-a-vis motivations for a purchase behaviour.

Like the basic tri-component model, theory of reasoned action also based on learn, feel and do components, however, they are arranged in a different pattern as:

The individual's attitude is measured as a sum of objective norm and subjective norm. Objective norms is measured as individual's belief for certain outcomes after a purchase and the importance of those outcomes.

Mathematically,

In accordance with this extended model, to understand intention we also need to measure the subjective norms that influence an individual’s intention to act. This can be assessed measuring a consumer’s feelings as to what means to the others (family, relatives, friends, co-workers). Here, the learning will basically measure emotional functions that may be accomplished through purchase. 

The individual’s motivation to comply with this normative belief is measured on the degree to which he is attributing his purchase to the others (family, relatives, friends, co-workers). This attribution different from person to person. The individual may feel more for the objects that has more value to the person most important for him for a particular purchase. 

Reference; Consumer Behaviour by Schiffman, Kaunk and Kumar and Consumer Behaviour by Solom

Mathematical approach to Multi-attribute attitude model with an illustration.

Extended Multi-attribute Attitude Model

Introduction to SPSS - Part 2

Adding Value Labels

Lets take a new variable as Gender. In your survey, you have marked two options - Male and Female. Let us code them as 1- male and 2 - female. These values are nominal. Place the following entries:

Name : Gen

Type: Numeric

Width : 1

Decimal : 0

Label: respondent gender

Measure: Nominal

Take the cursor to the cell for value. A blue ribbon is activated near 'none'. Click the blue ribbon. The following window appears. 

Type Value = 1 and Label = male.

Click Add

"1 = male" appears inside the dialogue box.

Similarly type Value = 2 and Label = female.

Click Add

"2 = female" appears in the dialogue box.

Press OK to save the Value Label.

The following screen will appear.

Now, Similarly define a variable Monthly Income.

Name: Incom

Type: Numeric

Width: 8

Decimal: 0

Label: monthly income

Value: 1 = less than 10000

2 = 10000 - 50000

3 = 50000 - 100000

4 = 100000 - 150000

5 = 150000 and above

(click on Add after defining each label, you can edit also using 'Change' and 'Remove')

Press OK

Change the measure to Scale as Income is a ratio scale variable.

Go to Data view. You can see your variables in the first row. 

When you fill the details for Gender and Income, you need to only put the code defined in Value Label.

If you want to see the actual value, activate the Value Label feature in the second row of task pane. For any analysis in SPSS, the codes are taken into consideration. Hence, deactivate Value Label (by clicking the same feature again) before any analysis.

Next : Importing an Excel file to SPSS

Introduction to SPSS Part 1

Entering Variables in a New File

SPSS or Statistical package for Social Sciences is a useful tool for analyzing data. It enables number of statistical operations. SPSS can be used for univariate, bivariate or multivarite data. For analyzing in SPSS, data can be exported from an excel file. Alternatively, data can also be directly entered in SPSS. Here is step be step process to define variables in SPSS:

1. Opening a New File

Upon opening SPSS following screen appears

Since we are entering variable in a new file, close the IBM SPSS Statistics 19 window. The following screen will appear. On the bottom of the screen look for the view type. Currently, we are in data view.

Click variable view, following screen will appear.

Now, you can enter the variable you want. 

2. Entering the variables.

Suppose the first variable is 'Name'. Type Name in the Name of the variable in the first row.

The other entries will be automatically filled.

Since 'Name' cannot be a numerical variable, the type needs to be changed. 

Click on the right side blue region near numeric, the following window will open.

Select String and Click OK. The Character space is 8, which can be increased, if required.

The following screen will appear. Now, our first variable is 'Name',  It's type is String, width is 8 character and it is of Nominal Scale (see Measurement & Scaling). In the label column, details about the variable is entered, like here, you can enter "respondent name". Now we shall enter second variable as 'Age'.

As we enter the variable it will automatically take other values. Since age is a discrete variable, we need to reduce decimal place from 2 to 0. For age the best measure is ratio scale variable, so we'll define the scale for 'Age'.

The drop down menu for measure shows three scaling types. select scale. Also the label can be defined as "respondent age"

So far we have defined two variables, one is nominal and the other is scale. The value is defined in case of variables where value labels are used.

Next: Using Value Labels for variables in SPSS

Testing of Hypothesis

The objective of this presentation is to explain hypothesis and the aim of hypothesis testing. The hypothesis is the backbone of any research. Hypothesis is important as it emphasize on researcher's ability to presume the solution. The job of the researcher is to put forward questions with its proposed answers.

The elaborate research process is eleven step process, where researcher deals with hypothesis twice. Basic hypothesis or supposition should be prepared after literature survey. The extensive study of the area of concern helps researcher identify the gaps, all the important factors to be studied and their relationships. All these acts as ingredients for the statement of hypothesis. This hypothesis is tested using statistical measures on data collected.

Hypothesis is the intuition of the researcher in terms of what are the probable solutions of the problem in hand. It should contain one dependent variable (criterion) and at least one independent variable (predictor). The aim of the hypothesis testing is to see the relationship between the two i.e. whether the predictor variable actually predicting about the criterion variable or not. Usually they are referred to as predictor and criterion variable in non experimental researches and independent and dependent variable only in experimental research.

In these examples, the hypothesis is stated in the form relationship. In the first example, attendance is the predictor variable, predicting marks of students. In the second example, performance is the predictor variable in judging two brands of automobile.

Hypothesis testing is the process that starts as early as in the step 3 of the process, with the statement of hypothesis and continuous till the penultimate step of the research process.

The Null Hypothesis is the statement of no difference. To begin with, researcher needs to assume that things shall not change because of something or anything. In typical marketing context, any effort (advertising, sales promotion, quality improvement, etc) taken to boost sales will not result in any increase in sales. In other words predictor variable has no control over criterion variable. Medicines will not sustain diseases, performances will not improve, are few examples of null hypothesis.

In some other cases, null hypothesis may state that all the predictor variable (if they are more than one) have equal response or no response to criterion variable. For example, all the factors like price, quality, availability, design, promotion are equally responsible for the success of a brand.

Alternate Hypothesis is the statement other than null. It highlights the differences, effects, and anomalies in the research area. It aims at proving anomalies, not stating the reason  behind them though. The aim of the research is prove alternate hypothesis and reject null. For a given null hypothesis there can be many alternate hypothesis.

In a given set observation, the parameters to represent a variable (criterion or predictor) are mean (X̅) and standard deviation (σ). In hypothesis testing the means of a sample (X̅) can be tested against the standard mean (µ) or two different samples can be studied ( µ1 and µ2)

The test of difference or no difference would have been be identified by merely looking at means values, but according to the concept of the deviation from mean more precisely, standard deviation (σ), the difference may be because of the extreme values in the observation set. Hence, it is important to see whether the difference in mean values is significant or it is because of the standard deviation. So, the difference is considered as no difference up to a level of significant difference. Since we are checking the probable answer in testing of hypothesis, the level of significance, denoted as α, is taken in terms of percentage probability. Usually, it is set at 5%, which means there is 5% probability that the researcher may state the difference when it is not significant. 

This initiates the discussion on errors in hypothesis testing. Type I error or α-error is committed, when the researcher rejects a null hypothesis when it was true. That means good product is rejected. Type II error or  β- error is caused when a null hypothesis is accepted when it was false. That mean accepting a bad product. Type I error is caused due to wrong selection of α. Researcher should have a reasonable level of significance, so that good samples are not rejected. But, in case if the researcher try to increase α he may accept bad samples also and may commit β- error.

It is evident that both types of errors can't be reduced simultaneously. There is  trade-off between two -types of error which means probability of making one type of error can only be reduced if the researcher is willing to increase the probability of making the other type of error.The appropriate level of α is decided examining the costs and penalties associated. If the task is to examine engine failures in aircraft it is advisable to keep a low α, but if they are paper planes, high α can be kept. In pharma industry, though, high α is recommended.

There are number of statistical tests available for testing of hypothesis. Depending upon the number of variables used one can select z-test, t-test,  χ2 test, F-test and so on.

Identification of the critical or rejection region using z-test taking α = 0.05 based on the central limit theorem that states that in a distribution all the values are equally distributed around the mean or in other words, they follow normal distribution.

Normal distribution is represented by bell shaped curve, at the middle of which the mean is located. According to null hypothesis, the hypothesized value, say X̅, is equal to the standard value i.e. µ. This is the state of no difference. If the value of X̅ is greater than µ, the z-value moves on the positive (right hand) side and if it is less, z-value moves on negative (left hand) side (see formula for z). Up to some point this more or less is not significant. But as per previously decided value of α , after some z-value the state of difference (more or less) becomes significant.

 The total area of normal curve is 1 (i.e. 100% probability). From the symmetry of normal curve, the area (probability) on each side is 0.50.

Two-tailed Test

When the researcher is only trying to identify the difference and not concerned about more or less, the level of significance is equally divided in both sides (tail) of the normal curve.This type of test is called as Two-tailed test. To determine the critical region, 0.025 (i.e.0.05/2) is subtracted from each side and the acceptance region remains to be 0.475 (i.e. 0.50 - 0.025) on each side.

This value when located in z-table, gives z value equals to 1.96

Again, from the property of normal curve z= -1.96 on left hand side and z=1.96 on the right hand side. hence for two-tailed test the acceptance region is between -1.96 to +1.96. For any value of z less than -1.96 or more than +1.96 null hypothesis will be rejected.

One-tailed Test

For identifying more or less situation, the whole 0.05 level of significance is allocated on one side of the normal curve. 

For the situation when the researcher wants to reject only the values that are less than standard value. Everything on the right hand side will be accepted and treated as no difference. For example, for testing the hypothesis if the overall performance of an institute has degraded or not, researcher is not interested in the parameters where it is improved, but only look for the parameters where there is a shortfall and aims at checking whether these shortfalls are significant or not. 

0.05 will be deducted from left side. the z value will be located for the area 0.45 (i.e. 0.50 - 0.05).

z-table does not have a value equals to 0.45. The two values that are equi-distant from 0.45 are 0.4495 and 0.4505. The z-value for 0.45 is taken as 1.645, the mean of the respective z-value of 0.4495 (i.e. 1.64) and 0.4505 (i.e.1.65). Hence the acceptance region is for anything more than -1.645 and the null hypothesis is rejected for any value of z less than -1.645.

On the other side of the tail, the researcher will only reject if the value is more than some standard value. Here everything on the left hand side is accepted as no difference and the rejection region will lie only on the right tail. For example, if a company wants to know whether the sales of a product is increased or not, it shall only see the higher sales figures for the significant increase.

Here, the null hypothesis is rejected if z value is more than 1.645 and accepted otherwise.

Hopefully, you have now a better understanding of hypothesis and testing of hypothesis. Later, I shall also discuss the alternate method of testing a hypothesis based on p value.