Multivariate data analysis

I work as an data analyst at a company that develops gas sensors. Gas sensors can be tricky and to get as good sensor as possible we are testing different treatments of the sensor component before it is produced.

During a test of a new treatment the sensor is tested for the most important criterias; accuracy and stability over time. This is done by having the sensors powered on under a long time period and to measure the performance of the sensors in some time points. The performance is measured by a number of variables.

We now want to analyse all test data that we have to investigate the impact of different process treatments (input parameters) on the performance of the sensors (output data). We want to find the dependencies of input and output data.

I have made some attempt to analyse the data by first convert the performance data over time to some parameters that summaries the variables over time, average, range etc. Then I have used PLS (with MATLAB) and some interesting things have been seen but it is hard to really catch the variation over time.

Does anyone know how to handle this type of problem. Is there any suitable technique to use?


TS Contributor
thois sounds like repeated measures anova. you have observations from a few time periods and see what happens. SPSS offers this possibility. you can also add covariates. Also another possibility is mixed effects model. it is doable in spss, but mainly through the syntax. R language has a ready function for that.