Rowise multiple chisquare test, selecting columns

[Lilip]

I have created the following df as example of my data. What I want to do is to perform chisquare test between pop1 (allele_count and allele_number) and pop2, for each SNP individually. Than the same test between pop1 and pop3. I have many SNPs and populations in real data.

I have tried rowise ( ), cause I want to perform the test by rows. But I I don´t know how to write the code to select 2 column , then another 2 columns and so forth.

Anyone could shed some light on that?

SNPs<-c("SNP_A", "SNP_B","SNP_C")
pop1_allele_count<-c(1,3,5)
pop1_allele_number<-c(100,100,100)
pop2_allele_count<-c(5, 7, 8)
pop2_allele_number<-c(100, 100, 100)
pop3_allele_count<-c(2, 3, 4)
pop3_allele_number<-c(100, 100, 100)
main<-data.frame(SNPs,pop1_allele_count,
pop1_allele_number,
pop2_allele_count,
pop2_allele_number,
pop3_allele_count,
pop3_allele_number)

[DavoWW]

Hi @Lilip,
Welcome to the RStudio Community Forum.

I noticed its 6 days since you posted your question. If you haven't solved it yet then this code may help:

suppressPackageStartupMessages(library(tidyverse))

# OP's data modified to make "pop3" values more extreme, for illustration
main <- data.frame(
  SNP = c("SNP_A", "SNP_B", "SNP_C"),
  pop1_allele_count = c(1, 3, 5),
  pop1_allele_number = c(100, 100, 100),
  pop2_allele_count = c(5, 7, 18),
  pop2_allele_number = c(100, 100, 100),
  pop3_allele_count = c(2, 3, 14),
  pop3_allele_number = c(100, 100, 100)
)

main
#>     SNP pop1_allele_count pop1_allele_number pop2_allele_count
#> 1 SNP_A                 1                100                 5
#> 2 SNP_B                 3                100                 7
#> 3 SNP_C                 5                100                18
#>   pop2_allele_number pop3_allele_count pop3_allele_number
#> 1                100                 2                100
#> 2                100                 3                100
#> 3                100                14                100

pivot_longer(main, cols=contains("pop")) %>%
  separate(., col=name, into=c("pop","allele","type")) -> tmp.df

pivot_wider(tmp.df, id_cols=c("SNP", "pop"), names_from="type") %>%
  mutate(absent = number - count) -> long.df

long.df
#> # A tibble: 9 x 5
#>   SNP   pop   count number absent
#>   <chr> <chr> <dbl>  <dbl>  <dbl>
#> 1 SNP_A pop1      1    100     99
#> 2 SNP_A pop2      5    100     95
#> 3 SNP_A pop3      2    100     98
#> 4 SNP_B pop1      3    100     97
#> 5 SNP_B pop2      7    100     93
#> 6 SNP_B pop3      3    100     97
#> 7 SNP_C pop1      5    100     95
#> 8 SNP_C pop2     18    100     82
#> 9 SNP_C pop3     14    100     86

long.df %>%
  group_by(SNP) %>%
  mutate(expected = mean(count)) %>%
  mutate(correction = ifelse(expected < 5, 0.5, 0)) %>% # Yates's correction
  mutate(chisq = (count - expected - correction)^2/expected) %>%
  print(.) %>%
  summarise(sum_chisq = sum(chisq),
            df = length(count)-1)
#> # A tibble: 9 x 8
#> # Groups:   SNP [3]
#>   SNP   pop   count number absent expected correction chisq
#>   <chr> <chr> <dbl>  <dbl>  <dbl>    <dbl>      <dbl> <dbl>
#> 1 SNP_A pop1      1    100     99     2.67        0.5 1.76 
#> 2 SNP_A pop2      5    100     95     2.67        0.5 1.26 
#> 3 SNP_A pop3      2    100     98     2.67        0.5 0.510
#> 4 SNP_B pop1      3    100     97     4.33        0.5 0.776
#> 5 SNP_B pop2      7    100     93     4.33        0.5 1.08 
#> 6 SNP_B pop3      3    100     97     4.33        0.5 0.776
#> 7 SNP_C pop1      5    100     95    12.3         0   4.36 
#> 8 SNP_C pop2     18    100     82    12.3         0   2.60 
#> 9 SNP_C pop3     14    100     86    12.3         0   0.225
#> # A tibble: 3 x 3
#>   SNP   sum_chisq    df
#>   <chr>     <dbl> <dbl>
#> 1 SNP_A      3.53     2
#> 2 SNP_B      2.63     2
#> 3 SNP_C      7.19     2

long.df %>%
  group_by(SNP) %>%
  summarise(stat = chisq.test(count)$statistic,
            df = chisq.test(count)$parameter,
            pval = chisq.test(count)$p.value,
            method = chisq.test(count)$method)
#> Warning in chisq.test(count): Chi-squared approximation may be incorrect

#> Warning in chisq.test(count): Chi-squared approximation may be incorrect

#> Warning in chisq.test(count): Chi-squared approximation may be incorrect

#> Warning in chisq.test(count): Chi-squared approximation may be incorrect

#> Warning in chisq.test(count): Chi-squared approximation may be incorrect

#> Warning in chisq.test(count): Chi-squared approximation may be incorrect

#> Warning in chisq.test(count): Chi-squared approximation may be incorrect

#> Warning in chisq.test(count): Chi-squared approximation may be incorrect
#> # A tibble: 3 x 5
#>   SNP    stat    df   pval method                                  
#>   <chr> <dbl> <dbl>  <dbl> <chr>                                   
#> 1 SNP_A  3.25     2 0.197  Chi-squared test for given probabilities
#> 2 SNP_B  2.46     2 0.292  Chi-squared test for given probabilities
#> 3 SNP_C  7.19     2 0.0275 Chi-squared test for given probabilities

^{Created on 2022-02-10 by the reprex package (v2.0.1)}

Hop e this helps.