[farahtahsin]

I am trying to find the factors that influence consumers attitude towards a cause. I have collected data from 2 countries and will be analysing the data seperately. I will finally compare the results. I used likert scale in the questionnaire.

I'm getting really weird result with correlation analysis. the alpha value isn't satisfactory either. I dont have enough time to do the survey again. Can I analyse the data using T-test?

I am stucked somebody help me!!

[Bradelygh]

No , t-test is for very small samples only n<20 , also can you be sure that your sample data is random and Representative of a parent population which can be modeled by a normal distribution?

[Gherkins]

What are the factors that you have measured? Can you provide a summary of all the variables you measured, please?

I'm not sure what you mean by the alpha level being unsatisfactory - if it's not significant, you may have to accept this and just report it as it is. It could be a consequence of not doing the right test, though - I'd need further details as stated above.

I do not know where Bradelygh's information about t-tests required N < 20 is from. I have never heard of this. If your N is > 20 then you can usually assume it will be normally distributed due to the Central Limit Theorem.

[sj27]

What exactly are you trying to correlate?

[Bradelygh]

(Original post by Gherkins)
What are the factors that you have measured? Can you provide a summary of all the variables you measured, please?

I'm not sure what you mean by the alpha level being unsatisfactory - if it's not significant, you may have to accept this and just report it as it is. It could be a consequence of not doing the right test, though - I'd need further details as stated above.

I do not know where Bradelygh's information about t-tests required N < 20 is from. I have never heard of this. If your N is > 20 then you can usually assume it will be normally distributed due to the Central Limit Theorem.

Its from a Statics 3 in the Further mathematics A-level, t-test can only be used on a sample size <20.

[sj27]

The convention is to use a t-distribution for n<20 and a normal distribution for greater than 20. A t-test is a statistical test, not requiring the sample size to be <20. I think bradelygh is confusing the two?

But I am not sure what the OP is trying to do. Is this for a master's thesis? I would have thought the idea was to plan what you were going to do with the data, it almost sounds like the OP hasn't got a result that suited her and is now just trying to find some statistical method that will make the data confess. Surely if the result is unexpected, that too is a finding and needs to be examined as to why? (disclaimer: my stats experience is in finance/econometrics, whereas this sounds like some other social science)

[farahtahsin]

(Original post by Gherkins)
What are the factors that you have measured? Can you provide a summary of all the variables you measured, please?

I'm not sure what you mean by the alpha level being unsatisfactory - if it's not significant, you may have to accept this and just report it as it is. It could be a consequence of not doing the right test, though - I'd need further details as stated above.

I do not know where Bradelygh's information about t-tests required N < 20 is from. I have never heard of this. If your N is > 20 then you can usually assume it will be normally distributed due to the Central Limit Theorem.

(Original post by sj27)
What exactly are you trying to correlate?

Actually I am trying to find the influence of cause reated marketing on consumers from 2 different country. These are my hypotheses and under each hypothesis I have several questions. I already have collected the data (116 sample from each country).

Dependent Variable
Consumers will be motivated to purchase products from companies involved in Cause Related Marketing.
1. In order to support a Cause Related Marketing program you are prepared to incur additional costs.
2. Sometimes you buy such products which are not necessary to you but buy products that you don't really need but do so only to support a particular cause.
3. You continue to buy the product if the price of the product increases and the money donated to the charity also increases by the same percentage.
4. You actively search for products related with a cause campaign charity while shopping.
Independent Variable:
H1: Positive attitude toward Cause Related Marketing is a reflection of positive attitude toward charitable activity.
1. You regularly donate money to support charities.
2. You will still purchase a product if the company supports the same charity you support on a regular basis.
H2: Consumers’ will have a high regard for the promoted brand because of company's Cause Related Marketing activity?
1. You are likely to purchase a brand which has an involvement with Cause Related Marketing rather than purchasing a brand which does not have any involvement with Cause Related Marketing.
2. It gives you a degree of satisfaction to purchase from a brand that supports a social cause.
3. When purchasing a new brand, you will choose a product that is some way involved in Cause Related Marketing.
4. You will change a brand if the company is not involved in any Cause Related Marketing.
5. Your perception towards the brand or the company changes if it is found to be involved in Cause Related Marketing.
H4: the amount donated by a company influences consumer perception of the Cause Related Marketing campaign and their intention to buy.
1. If the product costs 200 taka it is sufficient to donate 1 taka for the charity. (-ve question)
2. The company should donate a lump sum for the cause rather than donating a Percentage of the sales.
H5: People prefer to support local causes than an international cause.
1. You are likely to buy a product if the company donates a percentage of the sales for Ahsania British Heart Foundation
2. You are likely to buy a product if the company donates a percentage of the sales to feed the Starving children in Africa (-ve question)
H6: consumers will evaluate a Cause Related Marketing offer more positively if it supports a natural disaster rather than if it supports an ongoing cause
1. You are likely to purchase a product if the company supports a natural disaster (e.g. Helping Japan).
2. You are likely to purchase a product if the company supports an ongoing cause (. Donating to heart foundation, buying sports ecquipments for schools etc.) (-ve question)
3. The companies supporting disaster reliefs are more socially responsible than the companies supporting ongoing causes.
H7: Consumers will evaluate a Cause Related Marketing offer more positively if it’s a long term/frequent support rather than shorter term/infrequent support
1. It is better for a company to spend more money on the campaign for shorter period of time, rather than spending less money for longer time. (-ve question)
2. The companies that support a cause for longer time are more socially responsible rather than the companies that support a cause for shorter time.

My superviser is on vacation and I am stucked with my data I dont know what to do. I cant do correlation since theres not much internal consistency. I would appreciate your help.

[Gherkins]

OP, I am still rather confused about a) what you are measuring and b) what you expect to find. What you've listed provides your measures and your specific hypotheses, but not a clear summary of your variables or how you expect them to interact. Your hypotheses also seem to conflict with the way you've listed them as DV and IVs; for example, H5 suggests that likelihood to support a cause is a DV and the type of cause (local vs. international) is the IV.

What I suspect you are looking at is which of your IVs (positive attitude towards charities, regard for brand, amount donated, local/international cause, natural/ongoing disaster, long term/short term) is the most important predictor of your DV, motivation to purchase. For this you need multiple regression analysis, and the way you do it will depend on the specific nature of your variables - you say that they are all scores on Likert scales but some of your questions make them sound like categorical variables (yes/no), in which case you would need a logistic regression.

Does any of that help? I really think you need to speak to someone at your university who understands your project and its aims (my background is in psychology) - is there a stats person you could go and ask, or perhaps a colleague of your tutor?

[farahtahsin]

(Original post by Gherkins)
OP, I am still rather confused about a) what you are measuring and b) what you expect to find. What you've listed provides your measures and your specific hypotheses, but not a clear summary of your variables or how you expect them to interact. Your hypotheses also seem to conflict with the way you've listed them as DV and IVs; for example, H5 suggests that likelihood to support a cause is a DV and the type of cause (local vs. international) is the IV.

What I suspect you are looking at is which of your IVs (positive attitude towards charities, regard for brand, amount donated, local/international cause, natural/ongoing disaster, long term/short term) is the most important predictor of your DV, motivation to purchase. For this you need multiple regression analysis, and the way you do it will depend on the specific nature of your variables - you say that they are all scores on Likert scales but some of your questions make them sound like categorical variables (yes/no), in which case you would need a logistic regression.

Does any of that help? I really think you need to speak to someone at your university who understands your project and its aims (my background is in psychology) - is there a stats person you could go and ask, or perhaps a colleague of your tutor?

Hey! Thanks for the reply. Yes Actually When I made the questionnaire, I didnt think of any dependent/independent variable to be frank. The dependent variable I listed was also a part of my question. and my superviser actually approved the questionnaire. all the data are collected in likert scale and I talked to a GTA who told me that I cant analyse without having a dependent variable. I thought this one was most suitable to be a dependent variable. All I wanted to do was to prove the hypotheses. I don't know much SPSS and my supervisor is on vacation. I have to submit the report on 30th september but I am clueless of what to do. I know I sound like I am out of my mind. Don't know what to do with all the data.

PS: I did analyse the data but the alpha value is not over .5 for most of them Plus I asked a lot of negative questions in 3 of my hypotheses. Those are giving me negative result. A respondent might show positive attitude towards both international and local causes. In that case the result is negative. I coded the answers negatively so they are not consistent at all. I dont wanna wait till my sup. come back from vacation because I have so much to do. Please help!

[Gherkins]

(Original post by farahtahsin)
Hey! Thanks for the reply. Yes Actually When I made the questionnaire, I didnt think of any dependent/independent variable to be frank. The dependent variable I listed was also a part of my question. and my superviser actually approved the questionnaire. all the data are collected in likert scale and I talked to a GTA who told me that I cant analyse without having a dependent variable. I thought this one was most suitable to be a dependent variable. All I wanted to do was to prove the hypotheses. I don't know much SPSS and my supervisor is on vacation. I have to submit the report on 30th september but I am clueless of what to do. I know I sound like I am out of my mind. Don't know what to do with all the data.

You're welcome. It's absolutely vital that you work out how you expect your variables to interact - whether that's a dependent/independent variable relationship, a dependent/predictor(s) relationship, or just that two variables correlate together. Without knowing the background of your project it's hard for me to do that. I'm very surprised that your supervisor didn't help you identify how your variables would relate to one another, or prompt you to do so - I'd recommend paying another visit to the GTA if you are able to.

(Original post by farahtahsin)
PS: I did analyse the data but the alpha value is not over .5 for most of them

To be significant your p value needs to be less than .05, is this what you mean? The significance of your data, or lack thereof, is not the highest priority here, I feel. These are the findings that you have - if they're not significant, there's nothing you can do about it and it's not a reflection on your capability as a researcher. You just have to do the analysis properly and provide reasons why the results might not be significant. Null findings are extremely common, sadly. I read a while back that it takes 10 research notebooks' worth of work to publish one paper with significant findings.

(Original post by farahtahsin)
Plus I asked a lot of negative questions in 3 of my hypotheses. Those are giving me negative result. A respondent might show positive attitude towards both international and local causes. In that case the result is negative. I coded the answers negatively so they are not consistent at all. I dont wanna wait till my sup. come back from vacation because I have so much to do. Please help!

You need to to invert the scores from your negative items so that all items are in a positive direction. Take the raw score away from the maximum rating participants could give on the scale. Say the scale is 1 to 6, and the raw score on the negative question is 2. 6 - 2 = 4.

[sj27]

I think Gherkins is right - you need to talk to someone at your university who will have a clearer idea of what you are trying to do. (I'm not even sure why you are so hung up on the alpha - the betas are usually what one focuses on, at least in my field.). I really think it is best to find someone at the university rather than get advice from people whose stats knowledge has been applied to other fields, and moreover when you have no idea how qualified people are here to even answer in the first place. (edit for clarity- the last remark is not aimed at gherkins, just a general observation.)

[farahtahsin]

(Original post by Gherkins)
You're welcome. It's absolutely vital that you work out how you expect your variables to interact - whether that's a dependent/independent variable relationship, a dependent/predictor(s) relationship, or just that two variables correlate together. Without knowing the background of your project it's hard for me to do that. I'm very surprised that your supervisor didn't help you identify how your variables would relate to one another, or prompt you to do so - I'd recommend paying another visit to the GTA if you are able to.



To be significant your p value needs to be less than .05, is this what you mean? The significance of your data, or lack thereof, is not the highest priority here, I feel. These are the findings that you have - if they're not significant, there's nothing you can do about it and it's not a reflection on your capability as a researcher. You just have to do the analysis properly and provide reasons why the results might not be significant. Null findings are extremely common, sadly. I read a while back that it takes 10 research notebooks' worth of work to publish one paper with significant findings.



You need to to invert the scores from your negative items so that all items are in a positive direction. Take the raw score away from the maximum rating participants could give on the scale. Say the scale is 1 to 6, and the raw score on the negative question is 2. 6 - 2 = 4.

By alpha i meant Cronbach's Alpha which is used to find the correlation among questions under each hypothesis (derived by: Analyse-Scale-Reliability analysis in SPSS). I also found the The p values (Analyse-regression-linear in SPSS). For 1st 2 questions the p values are .000 and .001 which is less than .05. P value for others are over .05. I am worried because I have 2 or 3 questions under each hypothesis. Can it be the reason that my p value is over .05? Also, Can I test the hypotheses without the reliability test at first?

Thanks a lot for your help!

[farahtahsin]

(Original post by sj27)
I think Gherkins is right - you need to talk to someone at your university who will have a clearer idea of what you are trying to do. (I'm not even sure why you are so hung up on the alpha - the betas are usually what one focuses on, at least in my field.). I really think it is best to find someone at the university rather than get advice from people whose stats knowledge has been applied to other fields, and moreover when you have no idea how qualified people are here to even answer in the first place. (edit for clarity- the last remark is not aimed at gherkins, just a general observation.)

Thanks! I surely will go see the GTA again. But I am amazed seeing how people help each other here without even knowing them. I cant thank you guys enough for wanting to help me
XX

[sj27]

OK. Now see, I had never even heard of Cronbach's alpha until now (i just looked it up - never had to deal with the type of data it is used for), so that proves my point about advice here!!!!