Cfrgtkky

I have 3 variables, A1, A2 and A3

• 
  A1 is temperature


• 
  A2 is month


• 
  A3 is location

A2 has 2 months - March and May.A3 has 2 cities - Chennai and Dubai.

but when I do a correlation between A1 and A3:


cor(A1,A3, method = "pearson")
'y' must be numeric


How can I fix this, please?

Many Thanks,
Ishack

There are many ways to split the data, but the first question to answer is "what hypothesis do I wish to test?"

Here is example code using average daily high temperatures in Chennai and Dubai from timeanddate.com

# data collected from average high temperatures collected from 2005 - 2015

# https://www.timeanddate.com/weather/india/chennai/climate

# https://www.timeanddate.com/weather/united-arab-emirates/dubai/climate

rawData <- "

temperature,month,city

75,Jan,Dubai

78,Feb,Dubai

83,Mar,Dubai

92,Apr,Dubai

100,May,Dubai

103,Jun,Dubai

106,Jul,Dubai

107,Aug,Dubai

102,Sep,Dubai

96,Oct,Dubai

87,Nov,Dubai

79,Dec,Dubai

86,Jan,Chennai

89,Feb,Chennai

93,Mar,Chennai

97,Apr,Chennai

102,May,Chennai

100,Jun,Chennai

97,Jul,Chennai

95,Aug,Chennai

95,Sep,Chennai

92,Oct,Chennai

87,Nov,Chennai

86,Dec,Chennai"



tempData <- read.csv(text=rawData)



# t-test for average temperatures

t.test(tempData[tempData$city =="Dubai","temperature"],

       tempData[tempData$city == "Chennai","temperature"],

       paired=FALSE)



# paired t-test

t.test(tempData[tempData$city =="Dubai","temperature"],

       tempData[tempData$city == "Chennai","temperature"],

       paired=TRUE)

# correlation

cor(tempData[tempData$city =="Dubai","temperature"],

    tempData[tempData$city =="Chennai","temperature"])

Two Sample t-test

The two sample t-test tests the null hypothesis that the two means are equal, irrespective of the association between data collected between the two groups in the test. Sometimes the association between two groups may be based on time (as in the case of the temperature data), but the pairing may be based on other characteristics (e.g. twins in a study that has test and control groups where each pair of twins is randomly assigned to test and control groups).

> t.test(tempData[tempData$city =="Dubai","temperature"],

+        tempData[tempData$city == "Chennai","temperature"],

+        paired=FALSE)



    Welch Two Sample t-test



data:  tempData[tempData$city == "Dubai", "temperature"] and tempData[tempData$city == "Chennai", "temperature"]

t = -0.24817, df = 15.546, p-value = 0.8073

alternative hypothesis: true difference in means is not equal to 0

95 percent confidence interval:

 -8.765568  6.932235

sample estimates:

mean of x mean of y 

 92.33333  93.25000

Since 0 is within the 95% confidence interval, we accept the null hypothesis that there is no difference in the monthly average high temperatures between Chennai and Dubai.

Paired t-test

The paired t-test calculates the difference between the pairs of observations and tests the null hypothesis that the average difference is 0.

> # paired t-test

> t.test(tempData[tempData$city =="Dubai","temperature"],

+        tempData[tempData$city == "Chennai","temperature"],

+        paired=TRUE)



    Paired t-test



data:  tempData[tempData$city == "Dubai", "temperature"] and tempData[tempData$city == "Chennai", "temperature"]

t = -0.39555, df = 11, p-value = 0.7

alternative hypothesis: true difference in means is not equal to 0

95 percent confidence interval:

 -6.017343  4.184009

sample estimates:

mean of the differences 

             -0.9166667 

Since 0 is within the 95% confidence interval, we accept the null hypothesis that there is no difference in the monthly average high temperatures between Chennai and Dubai, when the test is conducted on the differences in pairs of monthly average high temperature values.

Correlation

The Pearson correlation measures the strength of the linear relationship between two variables, -1.0 = perfect negative correlation, 0 = no linear correlation, and 1 = perfect positive correlation.

> cor(tempData[tempData$city =="Dubai","temperature"],

+     tempData[tempData$city =="Chennai","temperature"])

[1] 0.7929018

> 

A correlation of 0.79 indicates a strong positive linear relationship between monthly average high temperatures in Dubai and Chennai.

Technique used to split the data

Since I created a raw data file and loaded it into R with read.csv(), I used the [ form of the extract operator to extract rows based on the value of the city column. I also created the raw data file in monthly order for each city, so the order of values in each subset matches by month, enabling a straightforward use of the pairwise t-test.

# extract temperature values for Dubai

tempData[tempData$city =="Dubai","temperature"]

A wide variety of techniques can be used to subset data from an R data frame, such as the which() function and the sqldf() function.