I have 80 variables containing measures of temperatures. At each hour, measures are recorded for all the 80 variables. As such the length of each dataset corresponding to each variables has the same length and there is no lag between measures for each variable at time t. Can I do a simple Pearson Correlation for these variables since each variable values are recorded at the same time t
hi
try with SPSS for pearson's correlation with variables and between the variables u can determine which is the most positively correlated and negatively correlated with the probablity of correlation at 50% and 100%
Kamal Nor
Yes, you will obtain an 80x80 correlation matrix, which is symmetrical and the main diagonal will contain only ones.
Similar to this: https://i.stack.imgur.com/xAmN5.png
Kamal Nor, Edgar Guevara
To me, before you apply Pearson's correlation on time-series, probably you need to consider the autocorrelation (aka "serial correlation") in your time-series.
Kindly see the random walks (coin flipping) example given by Gien_b at the following page: https://stats.stackexchange.com/questions/133155/how-to-use-pearson-correlation-correctly-with-time-series
If your time series have strong autocorrelation, then you will probably encounter the problem of the so-called "spurious correlation". The p-value of that comparison may also be inaccurate due to the miscalculation of the degree of freedom.
In your case, I would suggest you to think about your datasets first. Are your hourly-measured temperatures dependent on each other?. That is, for one variable, if you get a high temperature C = 37 degrees at t = 1, how likely are you also going to get a temperature around 37 degrees at t = 2? If your measures of temperatures are entirely independent of time and can vary freely from measure to measure, then you can apply Pearson's correlation directly and get the correct results.
- Badges
- Science topic