I need help , Problem with Regression analysis

[Juels mbevo]

Greetings dear readers,

i have a question
Ive made a survey with about 200 People. They answered 14 Questions with likert skalas but i have problem.
As you see every item is a variable and every person gave a scale from 1-5 on my likert skala (from never to always).
If i try to merge the Variables to make a dummy variables for a multiple regression analysis.
I wanted to see how the main Variable (viewers of Esport )is affected by the cognitive likert skala measured variable AND some control variable like age gender job.

I tried to make a dummy binary variable and then just using the regress
I created the binary variable from a likert skala ordinal variable.
I basically coded 0 == 1 and 1=!1. , because i messeaured their view preference by a likert skala. Is this a mistkae? Cannot i not just do 0 == 1 and 1=!1 , because it disturbes the regression command later on?

But if i try the regress command , stata does not respond. It just stops working and i have to break.
If i try areg it still does not work , it just doesnt give me output besides:
omitted because of collinearity

even if i make a regress command like: regress depend variable Gender -> it just does not give me output for over an hour.

Im using stata over vpn because im not at the university so thats maybe the problem and i have to drive there to do it.
But maybe you see some mistkaes

For that i uploaded pictures:
Picture 1+2= Dummy Variable and vEsportKog (summindex of likert skala for cognitive influences)

My goal is the see if the influences of vEsportKog is bigger then the influences of another Variable on the dummy variable.
So i want to do a multiple regression analysis.

I uploaded the files so maybe you can take a look and help me , i would be grateful!!

[Juels mbevo]

It works perfectly, i found a solution.

But i still have question, which command shall i use?
Regress or logit?

[Richard Williams]

Given that the original variable is ordinal, you could use ologit.

But, if for some reason you want to dichotomize, I would use logit.

[Juels mbevo]

if use the logit vEsportTVSport vGRes vGRtv Geschlecht Alter bildungserfolg Beruf command he says:
Note: 1 failure and 0 successes completely determined.

vGRes |
Trifft nicht zu | 0 (omitted)
Trifft eher nicht zu | 0 (empty)
ab und zu | 0 (empty)
Trifft eher zu | 0 (empty)
Trifft zu | 0 (empty)

note: 1.vGRes != 1 predicts failure perfectly
1.vGRes dropped and 160 obs not used

note: vGRtv != 1 predicts failure perfectly
vGRtv dropped and 6 obs not used

note: 2.vGRes omitted because of collinearity
note: 3.vGRes omitted because of collinearity
note: 4.vGRes omitted because of collinearity
note: 5.vGRes omitted because of collinearity



You can tell me why? Shall i then just go with the regression command instead of logit?
I read here that i dont need to make my depenend variable into a binary dummy.
https://www.researchgate.net/post/Wh...ert-scale_data
Because i measured the people that watch a or b by a likert skala.

So can you give me a command with which i can fuse those to likert variable watchers? Or is it simply summing them up and then dividing them through 2?
Because i have 2 likert Variables : One for Watching A and one for watching B
Because they had 5 options from 1 " i dont watch A at all" 2 "i watch it rarely" 3 "i watch it sometimes" ... etc until 5.
I basically wanted to exclude everyone from 1 " i dont watch it all" .

How can i do that so i can make them into one depended variable?

[Marcos Almeida]

You may wish to read about ‘perfect prediction’ and collinearity as well. These are important concepts when dealing with this sort of regression analysis.