hypothetis testing

freeze4real

Hi guys,

Is it possible to test the null hypothesis using the P-value ?

thanks

Ostrom

It is, but conceptually, you are not subjecting the null hypothesis to a 'test' as such. The null hypothesis is the logical opposite of the research hypothesis, which is what you are testing below (in your case, you are testing both the overall significance of the model, and that the coefficient of your independent variable - income - is different from 0).

Typically, a threshold of 0.05, (or 5%) is set, below which the 'null hypothesis' is rejected; therefore you need to establish from your own output, whether this threshold has been exceeded.

freeze4real

efla wrote: »

It is, but conceptually, you are not subjecting the null hypothesis to a 'test' as such. The null hypothesis is the logical opposite of the research hypothesis, which is what you are testing below (in your case, you are testing both the overall significance of the model, and that the coefficient of your independent variable - income - is different from 0).

Typically, a threshold of 0.05, (or 5%) is set, below which the 'null hypothesis' is rejected; therefore you need to establish from your own output, whether this threshold has been exceeded.

Hi efla,

From what I can deduce there is 2 methods of hypothesis testing here, research method and the null method or much thorough method.


With the P-value I cant test the null, My level of significance is 5% or 0.05.

So if the P-value is greater than level of significance I accept the null hypothesis.


With the other method, How do I compute it.

I have added a bunch of forumale but they seem vague to me.




Uploaded with ImageShack.us


thanks Elfa

Ostrom

There aren't two methods exactly - every research hypothesis (i.e. a hypothesized effect) has as its logical opposite the possibility of no effect (the null hypothesis). The specific formula varies depending on the nature the the data, method of analysis, and test statistic required. Those formulae are for computing ordinary least squares output by hand, which is quite tedious. You have all the pieces of information you need in your original post, however I suggest you start by reading through a basic test procedure such as the t-test to get a sense of how significance based hypothesis testing works. The p-value on the coefficient allows you to test the hypothesis that the effect (slope) of income is different from 0. If this value is less than .05, you typically accept the research hypothesis. However if it is greater, you 'cannot reject the null', and therefore conclude that income has no effect based on your data.

Also, if this is time series data, you will need to report the Durbin Watson test statistic (I'm assuming you need it since it is provided in your second post?), which has its own distribution, but the logic is the same.

ulyssesfcohen

The null hypothesis is usually the baseline against which we test test alternative hypotheses. It is the straw-man which states that there is no association between two variables, that they are completely unrelated. test of null hypothesis are relatively rare in social sciences - since some things are related in some form. So from theoretical standpoint, null hypothesis is rarely of interest. However, there are exceptions

To test the null hypothesis, you need to first do a power analysis (see Cohen, 1988)

Cohen, J. (1988). Statistical Power Test for the Behavioral Sciences. Lawrence Erlbaum, Hillsdale, NJ.