# Thread: Model for regression with multiple binomial predictors?

1. ## Re: Model for regression with multiple binomial predictors?

Greta, this is what I am trying to get at (ecological fallacy). There are 5 blue cars, 10 red cars, and 18 whites cars and a police traffic stop, 5 drivers were arrested for intoxication. You don't know which drivers were arrested. If it was random, then you can construct probabilities. But the fact that this whole question is being asked provides doubt in regards to it merrily being random. There are no numerators.

I got a bill at the restaurant for \$100 and I order two pizzas, cola, bread sticks. How much did each item cost? No idea. Now you do this over a bunch of restaurants, you may get a little hit of a pattern but the costs could be totally different at other restaurants (just like people using Google or direct URL can't assume to be the same each or across days). I think trying to use or justify a model still does not negate the overlying issue.

2. ## Re: Model for regression with multiple binomial predictors?

Thanks guys.

And guess what? That channel that didn't lead to sales? I've looked around Google Analytics all day and I've discovered an extensive bot network with servers all over the country. I just wrote up an 11-page report and sent it to my client. They've probably scammed my client out of thousands of euros worth of advertising, possibly tens of thousands. Wow. And I discovered all this just out of curiosity, thinking how I could tackle this statistical problem.

Just saying, statistics brings you places

3. ## Re: Model for regression with multiple binomial predictors?

Originally Posted by hlsmith
Greta, this is what I am trying to get at (ecological fallacy). There are 5 blue cars, 10 red cars, and 18 whites cars and a police traffic stop, 5 drivers were arrested for intoxication. You don't know which drivers were arrested. If it was random, then you can construct probabilities. But the fact that this whole question is being asked provides doubt in regards to it merrily being random. There are no numerators.

I got a bill at the restaurant for \$100 and I order two pizzas, cola, bread sticks. How much did each item cost? No idea. Now you do this over a bunch of restaurants, you may get a little hit of a pattern but the costs could be totally different at other restaurants (just like people using Google or direct URL can't assume to be the same each or across days). I think trying to use or justify a model still does not negate the overlying issue.
Ah, now I see what you mean. Yeah, I guess that's a fair point. I'm not sure how likely ecological fallacy type events are in this context, but that's certainly a possibility.