# Biostatistcs Data Analysis

Home Forums Methodspace discussion Biostatistcs Data Analysis

Viewing 7 posts - 1 through 7 (of 7 total)
• Author
Posts
• #3606

Hello everyone

I have  data related to a cohort study

i have one dependent variable which is (Disease Status) with tow valus, With Disease And Without Disease

and two independent variables the first one is Weight for six months, which are Quantitative reading

and the second independent variable is Diagnosis test with tow valus( Positive & Negative) for the same six months, which are  Qualitative reading.

My questions are:-

1- what is apropriate graph can i use to plot the relation between the dependent variable and independent variable with the six months, and what the steps to do that throug SPSS.

2- what the appropriate test should i use to show the risk of the independent variable on dependent variable

thank you

#3612

I am waiting for the help

#3611

You should construct a life table and draw a graph of the Kaplan-Meier survival estimates: http://en.wikipedia.org/wiki/Kaplan%E2%80%93Meier_estimator.

If observations are recorded monthly (e.g., 6 times in total), you should estimate a discrete hazard model to learn about the effect of your independent variable on the dependent variable. A discrete hazard model is often estimated with a logistic regression model on the probability of leaving the sample, where the sample is censored panel data (censoring means, in brief, tat people are set to missing whenever they die, i.e., leave the sample). You may benefit from reading this: http://data.princeton.edu/wws509/notes/c7s6.html. Note that in the logistic regression the z-statistic associated with the independent variable will be a test of the effect of your independent variable on your dependent variable. One issue, however, is that surviving individuals will display autocorrelation. If you do not correct for autocorrelation, your standard errors will often be too small, and hence your z-statistic will often reject the null even when the null is true. In Stata you can correct for autocorrelation with the option -cluster()- in the logistic command (-logit-).

If your time is recorded on a continuous basis, you should go for a cox regression.

/Kristian

#3610

Dear Hassabo,

you may run a scatter diagram to examine the relationship between the variables

Since your dependent variable is categorical, you can use the possiblity of logistic regression after considering the the second independent variablea as quantitative.you can also find  the appropriate test from SPSS

#3609
Jeremy Miles
Participant

1. I’m not sure what is meant by weight for 6 months.  Is that one value, or a series of values?

2. Logistic regression.

#3608

Thanks for everyone passed in my Topic

I want to clear the independent variables, as i told you i have two independent vars. weight & diagnostic

when my research contain 25 sample size or 25 opservations, so my data is like that

wieght1  wieght2 wieght3 …. wieght6     diagnostic1 diagonstic2 …. diagnostic6

1       22.4      20.4       18            12                yes             yes                 no

2       14         11          22            34                no               yes                yes

.

.

.

25      22        33         22              32              yes                no                  yes

#3607
Jeremy Miles
Participant

The most appropriate analysis sounds like a fixed effects regression, or possibly a random effects regression.  These are both fairly tricky though, and (I would doubt) you’re going to be able to do them from the help you get on a message board.

Why do you want to do this analysis?  If this is for a college assignment, or project, you need to find what is expected of you – your tutors probably won’t know what fixed effects regression is, and even if you manage to do it, you’ll need to explain it to them.  If this is because you work in the field, you probably need to get professional help.

Sorry to be gloomy.

Viewing 7 posts - 1 through 7 (of 7 total)
• You must be logged in to reply to this topic.