# how to build a central measure

#### WeeG

##### TS Contributor
Hi,

I need some advice, I got a series of question that my "cases" was asked to answer. these questions are YES/NO questions, coded 1/0. all the questions concern one's health ( was ill last year ? takes medications ?.... ). I want to build some health measure. how do I do that ?

usually I am used to have a series of ordinal variables, and then I use average to build a measure that reflects the person's opinion about X. but what do I do when I have for example 8 YES/NO questions, all dealing with the same big issue. I want to check for relationship between one's health ( which is being checked in the YES/NO questions ), and so other factor ( which is a single YES/NO question).

#### bugman

##### Super Moderator
Hey WeeG,

how about converting the yes/no answers in each group into proportions (%) data. You could then create "average" proportions and use confidence intervals to estimate population parameters....

Phil

#### WeeG

##### TS Contributor
just so I'll know I understand you...

if for example, for every case, I have 4 questions of YES/NO, then the data looks like this:

1 1 0 1 0
2 0 1 1 1
3 1 1 0 0
4 0 0 0 1
5 1 1 1 1

then you suggest to build the measure like this:

1 2/4
2 3/4
3 2/4
4 1/4
5 4/4

that will lead me to a number between 0 to 1, a real number. is it actually OK to use confidence intervals on it and to include it in statistical tests ?

cheers !

#### bugman

##### Super Moderator
yes, thats it.

Im not sure about performing stats on the small sample sizes, but you try try estimating CI through permutation.

#### WeeG

##### TS Contributor
the sample size ain't small, I just gave an example of 5 people, but actually I got many more, over 100

there is no limit of how many questions should be counted, minimal, right ?
do I have to check reliability somehow ? like I do with ordinal scale using cronbach's alpha ?

#### bugman

##### Super Moderator
WeeG,

sorry Im not familar with cronbach's alpha (commonly used in Psychology?).

Someone might want to jump in here...