comparing 2 means, but no t-test possible?

#1
Hi,
For my research I have participants who collected 'evidence'. Hereafter they recieved additional information, after which they could collect more 'evidence'. Some did collect more, and some did not. Now I want to analyse or there is a sign. difference between the point before they received information, and after they received information. I first did this with a paired sampled t-test. However, now my examiner (a statisticus) said this was not possible as my outcomes (after the extra information) would always be equal or more, than before the participatns received information. Now, i do not know which test I can use. Who can help me out?
 

Karabiner

TS Contributor
#2
What exactely do you want to test? Your examiner is right that logically
the second value can only be larger than the first value.

With kind regards

Karabiner
 
#3
What exactely do you want to test? Your examiner is right that logically
the second value can only be larger than the first value.

With kind regards

Karabiner
So I have two conditions. One that received extra information X (experimental condition) and one that received 'extra' information (control condition). All participants have to collect evidence (Part I) than they receive extra information (depending on the condition they are in). then the participants can collect more evidence (if they want) (Part II).

A analyzed my main data with a repeated measures ANOVA. where my two conditions where the between subjects ('group') factor and part I and part II the within-subjects ('time') factor.

Here comes the first problem. when I run the anova, my 'time*group' and 'time' results are exactly the same. However, I think this might be a result of that the participants in the control conditions, do not change anything. But I could be wrong.

The second problem is the following:
Only in the experimental condition group, I would like to see or there are significant differences between the data from part I and part II. I did this with a paired sampled t-test. My examiner said this was not possible because the second value can only be larger than the first value (I get that). But what I don't know is which test I should use then?

she said the following:
1. You could test or the difference is larger than zero.
2. you could look at the percentage of people that secured extra evidence, make an interval around this, and then test if 0 is in the interval.

Now as she is my examiner, she is not obligated to help me but can send 'tips' like point 1 and 2. Normally I would discuss this with my supervisor, but she also has no idea what my examiner would like to see. So I was hoping I could find the answer here.
 

Karabiner

TS Contributor
#4
So your dependent variable is amount of information collected during
Part 1 versus additional amount of information colected during part 2?
And in one of the conditions, participants collected no additional
information at all, during part 2?

My examiner said this was not possible because the second value can only be larger than the first value
So you dependent variable is total amount of information collected?
Say, in part 1 a participant collected 6 pices of information, in
part 2 she collected additional 2 pieces, and the values of the
dependent variable are 6 (part 1) and 8 (part 2) then?

With kind regards

Karabiner
 
#5
So your dependent variable is amount of information collected during
Part 1 versus additional amount of information colected during part 2?
And in one of the conditions, participants collected no additional
information at all, during part 2?


So you dependent variable is total amount of information collected?
Say, in part 1 a participant collected 6 pices of information, in
part 2 she collected additional 2 pieces, and the values of the
dependent variable are 6 (part 1) and 8 (part 2) then?

With kind regards

Karabiner
Hi Karabiner,

This exactly: Say, in part 1 a participant collected 6 pieces of information, in
part 2 she collected an additional 2 pieces, and the values of the
dependent variables are 6 (part 1) and 8 (part 2) then.

And in one of the conditions, participants collected no additional
information at all, during part 2. Exactly, they did receive extra information, but just none added any evidence in part II

In total I have 40 participants, 20 in each condition.
 

Karabiner

TS Contributor
#6
In that case I am not sure why you shouldn't carry out the 2 analyses as you already did.
The t-test analyses whether the amount of information collected by the experimental group
during part 1 ("amount A") differs from the amount collected during part 2 ("amount B"; though
I do not know whether this is an interesting question in the context of this study).
Maybe your supervisor got confused the same way as me here, since it was not clear whether
you wanted to compare amount A with amount B, or amount A with amount (A+B).

"you could look at the percentage of people that secured extra evidence, make an interval
around this, and then test if 0 is in the interval."
doesn't make sense to me, but I might
well be missing something here.

With kind regards

Karabiner
 
#7
In that case I am not sure why you shouldn't carry out the 2 analyses as you already did.
The t-test analyses whether the amount of information collected by the experimental group
during part 1 ("amount A") differs from the amount collected during part 2 ("amount B"; though
I do not know whether this is an interesting question in the context of this study).
Maybe your supervisor got confused the same way as me here, since it was not clear whether
you wanted to compare amount A with amount B, or amount A with amount (A+B).

"you could look at the percentage of people that secured extra evidence, make an interval
around this, and then test if 0 is in the interval."
doesn't make sense to me, but I might
well be missing something here.

With kind regards

Karabiner

I am slightly confused here. Because you say:
The t-test analyses whether the amount of information collected by the experimental group
during part 1 ("amount A") differs from the amount collected during part 2 ("amount B")

However, in my studydesign, I compared amount A with the amount (A+B).

"
in part 1 a participant collected 6 pieces (A) of information, in part 2 she collected an additional 2 pieces (B),
the values of the dependent variables are 6 (A) (part 1) and 8 (A+B) (part 2) then.

Or am in misinterpreting your post?
 

Karabiner

TS Contributor
#8
However, in my studydesign, I compared amount A with the amount (A+B).
Ok, so that's what we started from. It makes no sense to test here, since A+B must logically be >= A.
I thought you maybe had in mind to compare mean of A with mean of B.

Your examiner's suggestions 1 & 2 both do not make sense to me, but I might be missing
something crucial.

With kind regards

Karabiner
 
#9
Ok, so that's what we started from. It makes no sense to test here, since A+B must logically be >= A.
I thought you maybe had in mind to compare mean of A with mean of B.

Your examiner's suggestions 1 & 2 both do not make sense to me, but I might be missing
something crucial.

With kind regards

Karabiner
Oke, thank you that is clear then. Do you know how i could than analyze my data?