Correlating Numeric and Non Numerical Data

[joe_0838]

Hi Im Joe, I am trying to correlate 2 very different variables using the correlation function in excel data analysis tool. However my one variable is non numeric. To be specific I am trying to find out if the type of device that students use for online class(either a cellphone or a computer) correlates with their GPA. I start thinking of a way to make the device variable a numerical data so I did a thing like this 1-cellphone 2-computer and I put the 2 digit GPA of each respondent so it kinda looks like this
Device GPA
1 85
2 92
1 90
1 88
2 85
1 84

It actually worked but Im not really sure if my way of inputting my device variable is correct. I am not really familiar with Excel hope you guys help me :< Also Im not really sure with the input of their GPA I mean I can also put categories like from 1-5 scale where 1= 90-100 and 5= below. 75.

 

[diddi]

it is not possible to perform a meaningful correlation on this data set. you could determine mean and standard deviation for each group and prove they are statistically significantly different using a t-test

 

[joe_0838]

it is not possible to perform a meaningful correlation on this data set. you could determine mean and standard deviation for each group and prove they are statistically significantly different using a t-test

I see but how can I calculate for the device variable's mean and sd. The only data that I have for that variable is the respondents answer which is either a cellphone or a computer.

 

[diddi]

get all the scores for the cellphone people and find mean and sd, then same for other device

 

[joe_0838]

get all the scores for the cellphone people and find mean and sd, then same for other device

Thanks a lot

 

[joe_0838]

Im gonna use two tailed t test here w/ my alpha value as 0.05 right

 

[diddi]

yes 2 tailed gives you 2.5% of samples remaining at each end which is around the +- 2SD. as long as you are happy with that level of confidence. if the sample is large you could improve possibly.

 

[joe_0838]

yes 2 tailed gives you 2.5% of samples remaining at each end which is around the +- 2SD. as long as you are happy with that level of confidence. if the sample is large you could improve possibly.

Thanks a lot I have one last question. I have 40 respondents and they are randomly selected so I dont really know if the no. Of people who use cellphones are the same with the no. Of computer users my problem here is in almost all videos that I have seen who used t test ,they have equal no of data for example 20 scores for variable a and 20 for variable b. My plan is to get 20 cellphone users in different class in our department and 20 computer users. Can I still use t test with this kind of selection?

 

[joe_0838]

Thanks a lot I have one last question. I have 40 respondents and they are randomly selected so I dont really know if the no. Of people who use cellphones are the same with the no. Of computer users my problem here is in almost all videos that I have seen who used t test ,they have equal no of data for example 20 scores for variable a and 20 for variable b. My plan is to get 20 cellphone users in different class in our department and 20 computer users. Can I still use t test with this kind of selection?

Or will it be the same if my respondents are not equal for both variables like 18 cellphone users GPA amd 22 computer gpa

 

[diddi]

t test has (n -1) degrees of freedom ie it takes sample size into account. thats the beauty of it.