Dear Surajit Saha,

I do not think that a r of less that 0.1 is strong at all.

With a sample size of 101 and an r of that level, I think your data is poorly correlated, if none. 

With these data, and the results that you are presenting here, I find it difficult to conclude that your hypothesis in partially suported.

good luck!

Adri

With a larger sample size you may be able to have a :p

Dear Adrian Rodriguez,

Thanks for your suggestion.

Can I say CS4 is partially supported? As the value of r is 0.153 and p= 0.063 (“IF my HYPOTHESIS 2 is “Having high score on CS4 is suitable for performance of a person).

If you suggest any documents on where partially supported hypothesis was used it will be a great help for me.

With Regards,

Surajit

Based on the data shown, I would argue that this is a failed experiment (as already suggested by Adrian and Fang-Yong). However, I might try other analysis methods. Try a multivariate model/multiple regression approach. By including interaction terms you might find some hidden relationships. Please graph the data, graph the results, graph the residuals. Sometimes a simple transformation will improve the correlation between variables.

To get marginally significant, I let the threshold be 0.1. I cannot apply this selectively, and I cannot use 0.063 as a threshold, so I have four tests of which one is marginally significant. The probability of a type I error is 1-(1-alpha)^4 = 0.34. If I had left the p-value at 0.05, then I would get a probability of 0.19. It is too easy to argue that your marginally significant result is a type 1 error.

The simplest approach to data analysis would suggest that the experiment was a failure, but the simplest approach may be the real issue.

I agree with Timothy that this is a failed experiment. Data doesn't support your hypothesis. Try some other simple techniques.

Confidence in a relationship is formally determined not just by the correlation coefficient but also by the number of pairs in your data. If there are very few pairs then the coefficient needs to be very close to 1 or –1 for it to be deemed ‘statistically

significant’, but if there are many pairs then a coefficient closer to 0 can still be considered ‘highly significant’.

what is the sample size in your experiment? (how many pairs data set do you have)

scatterplotting your data (as already suggested above) may help you identify possible outliers responsible for watering down a potentially stronger association between performance and any of the 4 CS scores

the four independent variables (CS1, CS2, CS3 and CS4) and your dependent variable “performance” - are these continuous or categorical variables? If they are categorical - you shouldn't have used pearson correlation at all.

hello all,

i am analyzing some data and in which i got the p value = 0.1913, so it indicates it is not statistically significant, but i found the correlation r= 0.8075.

i have considered two variables and according to my results of r, it shows that both the variables are correlate to each other but at the other hand there is no significant p value.

so how can i interpret ?

hello all,

i want to ask, is it alright if my moderating variable has a positive correlation but not significant ?

No statistical significance simply means correlational hypothesis is not supported. Your data is not healthy.