Quantitative analysis

Home Forums Methodspace discussion Quantitative analysis

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
  • #3652

    Hey everyone, 

    i am very green when it comes to analysis. Could you please give me some guidance. My two variables under investigation are : mass media messages (independent variable) and public response (dependent). I am seeking to find out the effect of mass media messages on the public.

    Jeremy Miles

    We need to know a little more about how you are conceptualizing these things – are they continuous measures, or dichotomous?  And how you think the relationship might look like.


    It sounds very complicated – what are “mass media messages”?  Is that one thing, or 20 things?


    The very simplest thing to do if you had two simple variables would be to correlate them – say you wanted to compare age and height in kids, you could do a correlation, which would tell you how they were related.  But when I say ‘age’ everyone knows what I’m talking about, and there’s no argument how to measure it.  When you say “public response” this is less specific.


    Hope that helps a little.  Post more details and I (or someone else) might try to add some more.



    If i got your point i think you have fitt your two variables by Normality test, if your two variables are follow to Normal Distribution you can do Regretion Model to show the effect of your Independent variable on the Dependent variable

    becouse the correlation give you only the strength of the relationship.




    Jeremy Miles

    This wasn’t what I mean to say.    Correlation and regression are (essentially) the same test, so whether your data are normal isn’t relevant here.


    Can you give an example of some of your data?





    All the replies above have valid points. It is not possible without knowing nature and distribution of data to suggest as how to resolve the problem. I mean whether data are on categorical, ordinal, ratio scale or continuous or discrete and also how they are distributed. Further, the extent of reliability of public response is to be analysed and errors in responses are to be rectified. In selection of a method of analysis it is also important to specify pattern of relationship i.e., linear, logarithmic,  logit, curvilinear (type of curve among several possibles) as dictated by theory or guess (hypothesised by the researcher) etc. Though the nature of data is not known, even then it is advisable to plot both data arrays on a graph in accordance to ranks of values separately. Ranks of values should be graduated on x-axis and values should be scaled on y-axis. Ignoring outliers, inspection of this plot will give an idea how are the variables are distributed which may be transformed into almost normally distributed variables using a method most suited to the trend of data and a simple linear regression analysis may be carried out, if theory permits. Otherwise, construct a  scatter plot as is done in the case of independent and dependent variables. Inspection of scatter plot would give a good idea as how the two variables are related, linearly or otherwise. If theory or hypothesis under which you are investigating the problem allows you to postulate that the observed pattern of relationship is one possibility, select a technique and fit the model and see whether model and its parametres are statistically significant or not.  To test reliability of data of public responses, resurvey the same respondents with the same questions arranged in a different sequence. Establish correlation between earlier and response values/importance (rank) returned in resurvey. If correlation is significant, then multiply earlier response with the value of the correlation obtained in the case of a respondent. It would give credibility to the response value on the basis of reliability of the respondent. If correlation is statistically insignificant ignore that respondent altogether.  


    The first thing you should think about is how are measured your variables, and which are their indicators. If public response is dependent and was measured in nominal scale with two categories, you can undergo a logistic regression. Or even you can either make a simples association test to see if there’s any association. Not knowing the way you measure both variables makes me difficult to answer you. Just pls send me measuring scale and indicators, and first of all we may see how is the way to undergo analysis.

    There are some interesting pages with free open access to quantitative methods, as I work at health sciences, perhaps they have some biais to these disciplines, but may help you too, after I may be in conditions to assess you which are the tests to perform. Alejandra

    Just look at Internet for the stat pages/javastat


Viewing 6 posts - 1 through 6 (of 6 total)
  • You must be logged in to reply to this topic.