Hello everybody,
I aquired some data(x, y variables) at the different sites and time points. Can these data be analysed by regression analysis? I want to establish a relationship formula. However,I have read some paper which using the data at the same sites for different time points or at different sites for the same time point. Thus, I wonder if I can run the regression analysis of my data (see attachment).Look forward to your reply.
xuegang chen
Dear Xuegang Chen,
You can run a regression analysis on your data, no problem, but make sure which variable you want to predict. In your current setting, you predict median home price as a function of interest rate. Shouldn't it be the other way round, that interest rate depends on the price? Note also that a linear regression between these two variables will be independent of the year.
I would also test for normal distribution of the residuals, to make sure that the data do not violate the assumptions of linear regression. Otherwise you'd have to stick with a correlation chart or perform a non-linear model.
Best regards,
Thomas
Hi Thomas A Neubauer,
Thank you for your reply. In my data, the predicted variable is median home price. But I dont understand clearly what mean the sentence "what meaning a linear regression between these two variables will need be independent of the year" is. In addition, if the regression relationship between these two variables exist, what it indicate. Can the regression equation apply to predict the home price at site1 or site2 .....site12 at next year? What difference do there be between the result s, the data at the same sites for different time points and the data at different sites for the same time point? Look forward to your reply.
xuegang chen
What I meant is that only the two variables "interest" and "price" enter the linear regression, but neither year nor site is considered. Thus, you cannot predict the home price at a certain site but only at a certain interest in the linear model. If you want to test whether year matters, you'd have to add it as a (categorical) variable, for instance in a multiple regression. Or you use another linear regression to predict price as a function of year.
Best, Thomas
Thank you. Yes, I only use the two variables "interest" and "price" for the linear regression. I want to further explain the conditions of model. Firstly, the all locations belong to the A city. Secondly, I got the regression equation(Y=-23409+393349). Finally, I want to get the home price at the new site17 in 2004 year or at site1 in 2006 year. If it is impossible to do it, there are some other methods for the work. My data is in the attachfile.
Best regards,
xuegang chen
That's perfectly fine. You can fill in the interest rates for 2004 and 2006 as "x" into your regression equation to get the predicted values for price. I ran the analyses also with log10-transformed variables and the R² and p-value are even better (R² = 0.44), so the relationship seems to be not really linear. In that case the equation is: y = -1.3037x + 6.4746 (if you use the formula, mind to use log10-transformed interest rates as x and transform the resulting y back to "normal", i.e. 10^y).
Thomas
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining